CN104756116A - Safeguarding measures for a closed-loop insulin infusion system - Google Patents

Abstract

Processor-implemented methods of controlling an insulin infusion device for a user are provided here. A first method obtains a current insulin on board (IOB) value that estimates active insulin in the user, and compensates a calculated insulin infusion rate in response to the obtained IOB value. A second method supervises the operation of a glucose sensor by obtaining and processing insulin-delivered data and glucose sensor data for the user. An alert is generated if the second method determines that a current glucose sensor value has deviated from a predicted sensor glucose value by at least a threshold amount.

Description

For the safeguard procedures of closed-loop insulin infusion system

the cross reference of related application

This application claims the rights and interests of following patent file: the U.S. Provisional Patent Application the 61/694th that on August 30th, 2012 submits to, the U.S. Provisional Patent Application the 61/694th that on August 30th, No. 950 1 submits to, the U.S. Provisional Patent Application the 61/812nd that on April 17th, No. 961 1 submits to, the U.S. Patent application the 13/870th that on April 25th, No. 874 1 submits to, the U.S. Patent application the 13/870th that on April 25th, No. 902 1 submits to, the U.S. Patent application the 13/870th that on April 25th, No. 907 1 submits to, No. 910.The full content of above-mentioned these patented claims of quoting is incorporated to herein by reference at this.

Technical field

The embodiment of theme as herein described relates generally to drug delivery system, and more specifically, relates to the system for controlling infusion of insulin speed based on state variable feedback (state variable feedback).

Background technology

The pancreas of normal health subjects produces insulin and responds the rising of blood sugar level and insulin releasing is entered blood flow.The β cell (beta cell) be present in pancreas produces insulin and as required insulin secretion is entered blood flow.If β cell can not play a role or dead (a kind of situation being called type i diabetes) (or in some cases, if β cell produces insulin in shortage, type ii diabetes), so must provide insulin from another source to health.

Usually, because insulin can not be oral, so by injector to inject insulin.Recently, the use of Insulin Pump Therapy is increasing gradually, especially sends the insulin for diabetic.Such as, outside infusion pump is worn on belt, in pocket, etc., and this outside infusion pump is by being fed into insulin delivery in body with the infusion tube of percutaneous needles head or the conduit be positioned in hypodermis.Cut-off nineteen ninety-five, in the U.S., use infusion pump therapy less than the type i diabetes patient of 5%.At present, in the U.S., more than 900, in 000 type i diabetes patient more than 7% is using infusion pump therapy, and uses the number percent of the type i diabetes patient of infusion pump to increase with the absolute velocity per year over 2%.And the quantity of type i diabetes patient increases with the speed of annual 3% or higher.In addition, the type ii diabetes patient of increasing use insulin also uses infusion pump.Internist has recognized that the state of an illness of continuous infusion to diabetic provides better control, and internist also outputs continuous infusion prescription to patient more and more.Although pump therapy provides control, pump therapy can experience multiple complications, and this makes user not too want to use conventional external infusion pump.

In insulin pump, because pump makes insulin, characteristic changes, so use the quick acting insulin contrary with the insulin of the slower onset of injection to be very common.Because insulin company have developed the insulin of very fast onset, so the insulin of very fast onset is adopted usually very soon.But current pump is still subject to the restriction of the onset speed of the insulin used in this pump.

Summary of the invention

A kind of processor manner of execution is provided herein.Described method can be used for the insulin infusion devices controlling user.Some embodiments of described method relate to the operation of the processor structure with at least one processor device, thus obtain when insulin (IOB) value on header board, the estimated value of its representative to biologically active insulin in user's body.Described method is by being proceeded by calculating IOB speed based on obtained current I OB value at least in part by processor structure.Described method proceeds by determining the infusion of insulin speed after regulating based on calculated IOB speed and uncompensated infusion of insulin speed at least in part by processor structure.Described processor structure selects the final infusion of insulin speed of insulin infusion devices, and wherein, the infusion of insulin speed after determined adjustment, uncompensated infusion of insulin speed or current basal speed are selected as final infusion of insulin speed.

The present invention also provide a kind of control the insulin infusion devices of user by processor manner of execution.Some embodiments of described method start by producing current I OB value, the estimated value of described current I OB value representative to the biologically active insulin in user's body.Described method proceeds as follows: calculate IOB speed based on produced current I OB value at least partly, obtain uncompensated infusion of insulin speed, and determines the infusion of insulin speed after regulating according to following formula.

AdjustedRate(n)＝max(0；PIDRate(n)-IOBRate(n))

Described method proceeds by selecting final infusion of insulin speed according to following formula:

$\mathrm{FinalRate}\left(n\right)=\left\{\begin{array}{cc}\max (\mathrm{Basal};\mathrm{AdjustedRate}(n\left)\right),& \mathrm{PIDRate}>\mathrm{Basal}\\ \mathrm{PIDRate}\left(n\right),& \mathrm{PIDRate}\≤\mathrm{Basal}\end{array}\right.$

In this formula, AdjustedRate (n) is the infusion of insulin speed after determined adjustment; PIDRate (n) is the uncompensated infusion of insulin speed obtained; IOBRate (n) is the IOB speed calculated; FinalRate (n) is selected final infusion of insulin speed; And Basal is the current basal speed maintained by the insulin equipment of user.

The present invention also provides a kind of tangible and electronic storage medium with the executable instruction of processor of non-transitory, when the processor structure comprising at least one processor device performs described instruction, described instruction performs the method for the insulin infusion devices controlling user.In some embodiments, described method is started by estimated value current I OB value, and described current I OB value represents the amount of the biologically active insulin in user's body.Described method continues to carry out as follows: the current I OB value at least partly based on institute's estimated value calculates IOB speed; The infusion of insulin speed after regulating is determined at least partly based on calculated IOB speed and uncompensated infusion of insulin speed, and select the final infusion of insulin speed being used for insulin infusion devices, wherein, the infusion of insulin speed after determined adjustment, uncompensated infusion of insulin speed or current basal speed are selected as final infusion of insulin speed.Described method provides selected final infusion of insulin speed to regulate sending of insulin by insulin infusion devices subsequently.

The present invention also provides electronic equipment.Some embodiments of electronic equipment comprise the memory element that processor structure and at least one are associated with described processor structure.The executable instruction of described at least one memory element storage of processor, when processor structure performs described instruction, described instruction performs the method for the insulin infusion devices controlling user.Described method comprises the current I OB value of the amount calculating the biologically active insulin represented in user's body, IOB speed is calculated at least partly based on calculated IOB value, determine the infusion of insulin speed after regulating based on calculated IOB speed and uncompensated infusion of insulin speed at least partly, and select the final infusion of insulin speed being used for insulin infusion devices.Selection step selects infusion of insulin speed, uncompensated infusion of insulin speed or current basal speed after determined adjustment as described final infusion of insulin speed.

The present invention also provides a kind of electronic controller for insulin infusion devices.Described electronic controller comprises processor structure, and described processor structure comprises at least one processor device and at least one memory element be associated with described processor structure.The executable instruction of at least one memory element storage of processor described, when described processor structure performs described instruction, described instruction provides IOB compensating module, this module estimated value represents the IOB value of the amount of the biologically active insulin in user's body, calculate IOB speed based on institute estimated value current I OB value at least partly, and determine the infusion of insulin speed after regulating based on calculated IOB speed and uncompensated infusion of insulin speed at least partly.IOB compensating module selects the final infusion of insulin speed of insulin infusion devices, and wherein, described final infusion of insulin speed is selected as the infusion of insulin speed after determined adjustment, uncompensated infusion of insulin speed or current basal speed.IOB compensating module provides selected final infusion of insulin speed to regulate sending of insulin by insulin infusion devices subsequently.

The present invention also provides the exemplary embodiment of electronic equipment.Described electronic equipment comprises the processor structure with at least one processor device, and at least one memory element be associated with described processor structure.The executable instruction of at least one memory element storage of processor described, when described processor structure performs described instruction, described instruction performs the method for the insulin infusion devices controlling user.Described method runs insulin equipment in the closed loop mode so that insulin delivery is delivered to user's body, obtain the current insulin delivering data of the amount representing the insulin sent by insulin infusion devices in the nearest sampling period, obtain the current sensor data representing the user's current sensor dextrose equivalent corresponding to the nearest sampling period, and historical insulin delivering data and the historical sensor data of the multiple history samples time periods of process before nearest sampling time section continue, thus obtain the sensor dextrose equivalent in the historical time cycle of prediction.Described method proceeds as follows: the difference between the current sensor dextrose equivalent of the nearest sampling time section of calculating current sensor dextrose equivalent and prediction, wherein, the sensor dextrose equivalent in the historical time cycle of prediction comprises the current sensor dextrose equivalent of prediction.Described method proceeds by giving the alarm when described difference exceedes threshold error amount.

Detailed description hereafter also relates to the tangible and electronic storage medium with the executable instruction of processor of non-transitory, when the processor structure comprising at least one processor device performs described instruction, described instruction performs the method for the insulin infusion devices controlling user.Described method comprises runs insulin infusion devices in the closed loop mode so that insulin delivery is delivered to user's body.Described method proceeds as follows: train the baseline historical sensor dextrose equivalent obtained in sampling process being identified in from the historical sensor dextrose equivalent of user.Described method calculates the stand-by scheme of multiple sensor glucose predictions model, wherein, each in described multiple stand-by scheme is calculated as the function of the historical insulin delivering data of bounded starting condition and user, and wherein, described bounded starting condition is subject to the impact of baseline sensor dextrose equivalent.Described method proceeds as follows: from calculated multiple stand-by scheme, select optimum matching scheme with comparing of the Part I of historical sensor dextrose equivalent based on from the sensor dextrose equivalent of prediction in the multiple stand-by scheme calculated.The sensor dextrose equivalent of the prediction in optimum matching scheme compares with the Part II of historical sensor dextrose equivalent, wherein, the Part I of described historical sensor dextrose equivalent corresponds to history cycle remote, the Part II of described historical sensor dextrose equivalent corresponds to nearest history cycle, and history cycle remote is before carrying out the nearest history cycle of data sampling.Described method by response described compare described historical sensor dextrose equivalent Part II off-target matching scheme at least threshold error value time give the alarm and proceed.

The present invention is also provided for the embodiment of the electronic controller of insulin infusion devices.At least one memory element that described electronic controller comprises processor structure and is associated with described processor structure, described processor structure comprises at least one processor device.The executable instruction of at least one memory element storage of processor described, when described processor structure performs described instruction, described instruction supplies a model manager module, thus insulin delivering data and current sensor data is obtained in insulin equipment operation with closed ring process, described insulin sends the amount of the insulin that insulin infusion devices is sent in the data Biao Shi nearest sampling period, and described current sensor data represents the current sensor dextrose equivalent of the user corresponding to the nearest sampling period.Model manager module define model training phase in the historical time cycle and model prediction phase and found relative to the interim acquisition of model training historical sensor dextrose equivalent with the scheme of sensor glucose predictions model optimum matching, wherein, described optimum matching scheme is the function of the baseline sensor dextrose equivalent obtained in the model training cycle, and is the function of the user's historical insulin delivering data obtained in the historical time cycle.The sensor dextrose equivalent of at least one prediction in optimum matching scheme compares with at least one the historical sensor dextrose equivalent only corresponding to the model prediction cycle by described model manager module, and response ratio is comparatively, sensor dextrose equivalent that at least one historical sensor dextrose equivalent described departs from least one prediction described at least threshold error value time give the alarm.

Detailed description hereafter also comprises the processor manner of execution of the insulin infusion devices controlling user.Described method starts so that insulin delivery is delivered to user's body by running insulin infusion devices in the closed loop mode.Described method proceeds as follows: the current insulin delivering data obtaining the amount representing the insulin that insulin infusion devices is sent in the nearest sampling period, obtains the current sensor data of expression corresponding to the current sensor dextrose equivalent of the user in nearest sampling period and the pre-treatment historical insulin delivering data in the nearest sampling period and historical sensor data and continues multiple history samples cycle to obtain the sensor dextrose equivalent of the prediction in historical time cycle.Described method calculates the difference between the current sensor dextrose equivalent of current sensor dextrose equivalent and prediction in the nearest sampling period subsequently, and wherein, the sensor dextrose equivalent of the prediction in the historical time cycle comprises the current sensor dextrose equivalent of prediction.Give the alarm when difference exceedes threshold error value.

Detailed description hereafter also comprises the processor manner of execution of the insulin infusion devices controlling user.Described method starts to user's body with insulin delivery by running insulin infusion devices in the closed loop mode.Described method proceeds by training the baseline historical sensor dextrose equivalent that obtains in sampling process identifying from the historical sensor dextrose equivalent of user.Next, the multiple stand-by scheme of calculating sensor glucose predictions model, wherein, each in described multiple stand-by scheme is calculated as the function of bounded starting condition and user's historical insulin delivering data, and wherein, bounded starting condition is subject to the impact of baseline sensor dextrose equivalent.Described method by selecting optimum matching scheme to proceed with comparing of the Part I of historical sensor dextrose equivalent based on the sensor dextrose equivalent of prediction from the multiple stand-by scheme calculated from the multiple stand-by scheme calculated.The sensor dextrose equivalent of at least one prediction in optimum matching scheme compares with the Part II of historical sensor dextrose equivalent, wherein, the Part I of described historical sensor dextrose equivalent corresponds to history cycle remote, the Part II of described historical sensor dextrose equivalent corresponds to nearest history cycle, and history cycle remote occurred before the nearest history cycle of data sampling.The Part II of described historical sensor dextrose equivalent depart from described optimum matching scheme at least threshold error value time, response ratio comparatively gives the alarm.

Another embodiment of the processor manner of execution of the insulin infusion devices controlling user is hereafter also provided.Described method comprises runs insulin infusion devices in the closed loop mode so that insulin delivery is delivered to user's body, define model training cycle in the historical time cycle and model prediction cycle and find the optimum matching scheme of the sensor glucose predictions model relative to the historical sensor dextrose equivalent obtained in the model training cycle, wherein, described optimum matching scheme is the function of the baseline sensor dextrose equivalent obtained in the described model training cycle, and is the function of the historical insulin delivering data obtained in the historical time cycle.Described method is by comparing the sensor dextrose equivalent of at least one prediction in optimum matching scheme and only proceeding corresponding at least one historical sensor dextrose equivalent in model prediction cycle.At least one historical sensor dextrose equivalent depart from least one prediction sensor dextrose equivalent at least threshold error value time, response ratio comparatively gives the alarm.

Foregoing invention content provides the introduction to selected concept in the mode briefly introduced, and selected concept is hereafter being described in detail.Foregoing invention content had both been not intended to the key characteristic or the essential characteristics that show the theme that application claims is protected, was also not intended to the supplementary mode as the scope determining the theme that application claims is protected.

Accompanying drawing explanation

By reference to describing in detail and claim to the more complete understanding of theme of the present invention generation, in following accompanying drawing, identical element can be referred at the identical Reference numeral of whole accompanying drawing in conjunction with following accompanying drawing.

Fig. 1 is the block diagram of closed loop glucose control system according to the embodiment of the present invention.

Fig. 2 is the front elevation of the closed loop hardware be positioned at according to the embodiment of the present invention in main body.

Fig. 3 A is the skeleton view of the glucose sensor system for embodiments of the present invention.

Fig. 3 B is the cross-sectional side view of the glucose sensor system of Fig. 3 A.

Fig. 3 C is the skeleton view of the sensor stand of the glucose sensor system of Fig. 3 A for embodiments of the present invention.

Fig. 3 D is the cross-sectional side view of the sensor stand of Fig. 3 C.

Fig. 4 is the cross-sectional view of the test side of the sensor of Fig. 3 D.

Fig. 5 is the vertical view for the infusion apparatus with fluid reservoir passage under the open mode of embodiments of the present invention.

Fig. 6 is the side view of the infusion assembly with the insertion syringe needle extracted for embodiments of the present invention.

Fig. 7 is the circuit diagram of sensor according to the embodiment of the present invention and power supply supply thereof.

Fig. 8 A is the schematic diagram of individual equipment according to the embodiment of the present invention and assembly thereof.

Fig. 8 B is the schematic diagram of two equipment according to the embodiment of the present invention and assembly thereof.

Fig. 8 C is another schematic diagram of two equipment according to the embodiment of the present invention and assembly thereof.

Fig. 8 D is the schematic diagram of three equipment according to the embodiment of the present invention and assembly thereof.

Fig. 9 lists the equipment of Fig. 8 A to Fig. 8 D and the form of assembly thereof.

Figure 10 is the block diagram of the glucose sensor system of Fig. 3 A.

Figure 11 A is according to the embodiment of the present invention for the details block diagram of the A/D converter of the glucose sensor system of Figure 10.

Figure 11 B is according to the embodiment of the present invention for the details block diagram of the A/D converter of the glucose sensor system of the Figure 10 with pulse persistance output intent option.

Figure 12 is according to the embodiment of the present invention with the circuit diagram of the I-FA/D converter of Figure 10 of node signal chart.

Figure 13 is according to the embodiment of the present invention with another circuit diagram of the I-F A/D converter of Figure 10 of the chart of node signal.

Figure 14 is according to the embodiment of the present invention with the another circuit diagram of the I-F A/D converter of Figure 10 of the chart of node signal.

Figure 15 is the circuit diagram of the I-V A/D converter of Figure 10 according to the embodiment of the present invention.

Figure 16 is the block diagram of the glucose sensor system of Figure 10 with prefilter and wave filter according to the embodiment of the present invention.

Figure 17 is the example chart of prefilter of Figure 16 according to the embodiment of the present invention and the chart of the effect to digital sensor values Dsig thereof.

Figure 18 is the frequency response chart of the wave filter of Figure 16 according to the embodiment of the present invention.

Figure 19 A is the sensor signal and the time dependent curve of unfiltered sensor signal of filtering according to the embodiment of the present invention.

Figure 19 B is the close-up section of the curve of Figure 19 A according to the embodiment of the present invention.

Figure 20 be according to the embodiment of the present invention be connected to the sensor stand of human body and the cross-sectional view of infusion assembly.

Figure 21 is the frequency response chart of the S filter that time delay according to the embodiment of the present invention corrects.

Figure 22 carries out the time dependent curve of digital sensor value Dsig before and after time delay correction relative to actual glucose measured value according to the embodiment of the present invention.

Figure 23 A is the schematic diagram of glucose clamp experiment (glucose level over time).

Figure 23 B is the curve of the insulin level in normal glucose tolerance (NGT) individuality of the various different level of the glucose clamp experiment of response diagram 23A.

Figure 24 A is the schematic diagram of glucose clamp experiment.

Figure 24 B is the schematic diagram of the proportional insulin of the glucose clamp experiment of response diagram 24A according to the embodiment of the present invention.

Figure 24 C is the insulin integrogram of the glucose clamp experiment of response diagram 24A according to the embodiment of the present invention.

Figure 24 D is the insulin derivative figure of the glucose clamp experiment of response diagram 24A according to the embodiment of the present invention.

Figure 24 E be proportional, the integration of the merging of the glucose clamp experiment of response diagram 24A according to the embodiment of the present invention with the insulin schematic diagram of derivative.

Figure 25 A is the insulin curve of the individuality of responsive movement training and the glucose clamp experiment of normal individual.

Figure 25 B is that histogram taken in by training glucose that is individual and normal individual.

Figure 26 feds back through based on glucose level the block diagram that infusion of insulin controls the closed-loop system of blood sugar level according to the embodiment of the present invention.

Figure 27 is the details block diagram being positioned at a part for the control loop of Figure 26 of body according to the embodiment of the present invention.

Figure 28 A and Figure 28 B are the curves for the insulin response measured by different normal glucose tolerance (NGT) individuality of two groups of glucose clamp experiment of embodiments of the present invention.

Figure 29 A be according to the embodiment of the present invention in glucose clamp experiment process relative to the curve of the output of two of glucose metering reading groups of different glucose sensors.

Figure 29 B is the curve of the actual insulin concentration in the blood of the insulin concentration that the glucose clamp experiment of response diagram 29A according to the embodiment of the present invention controls relative to controller.

Figure 30 is according to the embodiment of the present invention for the vertical view of the end of the Multifunction Sensor of measure glucose concentration and pH.

Figure 31 A is according to the embodiment of the present invention relative to the blood sugar schematic diagram of the sensor blood sugar of measure of the change in time.

Figure 31 B is the schematic diagram of the same time section inner sensor sensitivity according to the embodiment of the present invention shown in Figure 31 A.

Figure 31 C is the schematic diagram of the same time section inner sensor resistance according to the embodiment of the present invention shown in Figure 31 A.

Figure 32 is the block diagram using the derivative of sensor resistance to determine when again correcting sensor or more emat sensor according to the embodiment of the present invention.

Figure 33 A is the time dependent curve of analog sensor signal Isig according to the embodiment of the present invention.

Figure 33 B is the curve of the identical time period inner sensor resistance according to the embodiment of the present invention shown in Figure 32 A.

Figure 33 C is the curve map of the derivative of the sensor resistance of Figure 32 B according to the embodiment of the present invention.

Figure 34 A is the upward view of remote measurement characteristic monitor according to the embodiment of the present invention.

Figure 34 B is the upward view of different remote measurement characteristic monitor according to the embodiment of the present invention.

Figure 35 A is the plasma insulin sketch map of the glucose clamp responded according to the embodiment of the present invention in normal glucose tolerance (NGT) individuality.

Figure 35 B be according to the embodiment of the present invention when be delivered to hypodermic insulin by be directly delivered to be delayed in blood flow time Figure 35 A plasma insulin response schematic diagram.

Figure 36 A is the time dependent schematic diagram of plasma insulin concentrations after insulin heavy dose is directly delivered to blood flow according to the embodiment of the present invention.

Figure 36 B is the time dependent schematic diagram of plasma insulin concentrations after insulin heavy dose is delivered to hypodermis according to the embodiment of the present invention.

Figure 37 is the block diagram of the closed-loop system of the Figure 26 adding controller post-equalization device and derivative filter according to the embodiment of the present invention.

Figure 38 A is sensor signal measured value and the time dependent curve of Via measured value according to the embodiment of the present invention.

Figure 38 B be measure according to the embodiment of the present invention to the time dependent curve of electrode voltage Vcnt.

Figure 38 C is the time dependent curve of transducer sensitivity calculated according to the embodiment of the present invention.

Figure 38 D is the calculated value R of sensor resistance according to the embodiment of the present invention _s1time dependent curve.

Figure 38 E is another calculated value R of sensor resistance according to the embodiment of the present invention _s2time dependent curve.

Figure 38 F is the sensor resistance R of Figure 38 D according to the embodiment of the present invention _s1the time dependent curve of derivative.

Figure 38 G is the sensor resistance R of Figure 38 E according to the embodiment of the present invention _s2the time dependent curve of derivative.

Figure 38 H is curve when sensor is changed in time according to the embodiment of the present invention.

Figure 39 A and Figure 39 B is the block diagram of closed loop glucose control system according to the embodiment of the present invention.

Figure 40 is the block diagram of automatic blood extraction and re-injection according to the embodiment of the present invention.

Figure 41 A is actual blood glucose concentration curve according to the embodiment of the present invention.

Figure 41 B is the curve of actual insulin concentration in the blood of the insulin concentration that the blood sugar according to the embodiment of the present invention in response diagram 41A controls relative to controller.

Figure 42 illustrates the control feedback block diagram of state change feedback according to the embodiment of the present invention.

Figure 43 is the time dependent curve of basal insulin delivery rate using different ride gains according to the embodiment of the present invention.

Figure 44 uses the time dependent curve of the subcutaneous insulin of different ride gains according to the embodiment of the present invention.

Figure 45 uses the time dependent curve of the plasma insulin of different ride gains according to the embodiment of the present invention.

Figure 46 is the time dependent curve of insulin effect using different ride gains according to the embodiment of the present invention.

Figure 47 is the time dependent curve of simulation concentration of glucose using the PID controller fed back with state change and the PID controller using not carrier state change feedback according to the embodiment of the present invention.

Figure 48 uses to send time dependent curve with the PID controller of state change feedback and the simulation insulin of PID controller that carrier state change is not fed back according to the embodiment of the present invention.

Figure 49 is the processing module of exemplary embodiment and the block diagram of algorithm that illustrate closed-loop system controller.

Figure 50 is the process flow diagram of the exemplary embodiment of the control procedure illustrating insulin infusion devices.

Figure 51 is the figure of integration limits value relative to sensor glucose level.

Figure 52 is the block diagram that illustrated exemplary illustrates the exemplary embodiment of insulin (IOB) compensating module on plate.

Figure 53 is the process flow diagram of the illustrative embodiments illustrating IOB compensation process.

Figure 54 is the figure describing some time-event relevant with the operation of mode manager module.

Figure 55 is the process flow diagram of the exemplary embodiment illustrating mode sensor management process.

Figure 56 is the process flow diagram of the exemplary embodiment illustrating mode sensor training process, and the mode sensor management process that described mode sensor training process can describe with Figure 55 is combined and performs.

Figure 57 is the figure illustrating the malfunction that two kinds of can be detected by mode manager module are exemplary.

Embodiment

Following detailed description is only illustrate and be not intended to limit the embodiment of present subject matter or the application of these embodiments and purposes in essence.Word used herein " example " refers to " as example, example or illustrate ".Any embodiment exemplarily described herein might not be construed as the preferred or favourable embodiment relative to other embodiments.And, be not intended to herein to be subject in first technical field, background, the restriction of any theory expressed or imply that briefly introduces or exist in following detailed description.

The symbol of the operation that referential expression can be performed by various different computing element or equipment, Processing tasks and function, can according to function element and/or logical block elements description technique and technique in this article.That that these operations, task and function perform sometimes referred to as computing machine, computerized, software performs or computing machine execution.Should be understood that, the various different frame element shown in accompanying drawing by any amount of hardware, software and/or can be configured to the firmware components identification performing specific function.Such as, the embodiment of system or element can use various different integrated circuit component, such as, and memory element, digital signal processing element, logic element, look-up table, etc., they can a kind of or more than the control of a kind of microprocessor or other opertaing devices under perform several functions.

When implementing in software or firmware, the various different element of system as herein described is the code segment or the order that perform various different task in essence.Program or code segment can be stored in any tangible and in the processor readable medium of non-transitory." processor readable medium " or " machine readable media " can comprise any storage or the medium of transitional information.The example of processor readable medium comprises circuit, semiconductor memory apparatus, ROM, flash memory, erasable ROM (EROM), floppy disk, CD-ROM, CD, hard disk, etc.

Various different tasks performed by relevant from process as herein described can be performed by software, hardware, firmware or their any combination.Should be understood that, described process can comprise any amount of additionally or optional task, this task shows with special pattern, do not need to perform with exemplified order, and described process can be incorporated in the extra functional more comprehensive program or process having and do not describe in detail herein.And, a kind of in task shown in the drawings or can omit from the embodiment of described process more than a kind of, as long as the overall function wanted keeps complete.

As shown in illustrational accompanying drawing, the present invention implements in closed loop infusion system, and this closed loop infusion system enters the liquid stream infusion velocity of user's health based on the feedback regulation of the analyte concentration measurement value taking from health.In specific embodiment, the present invention implements in the controls, and this control system regulates based on the glucose concentration measurement taking from health the infusion of insulin speed entering user's health.In a preferred embodiment, described system is designed to simulation pancreatic beta cell (beta cell).In other words, the concentration curve uelralante that described system controls concentration curve that infusion apparatus sets up with the people's beta cell perfected with function similar when responding blood sugar concentration change in body enters user's health.

Therefore, described system simulation human body is to the natural insulin response of blood sugar level, and it not only effectively uses insulin, but also is responsible for other body functions, because insulin has metabolism and Mitosis.But the necessary accurate simulation beta cell of algorithm, does not consider that because being designed to minimize glucose oscillation in body the algorithm of insulin delivering amount can cause excessive body weight increase, hypertension and arteriosclerosis.In a preferred embodiment of the invention, in described system parody insulin secretion pattern and in the body regulating this pattern and normal health subjects to experience beta cell change consistent.Insulin sensitivity (the S there is normal glucose tolerance (NGT), significantly changing _i) experimenter body in beta cell reaction be maintain the best insulin response of glucose homeostasis.

As shown in Figure 1, glucose sensor system 10, controller 12 and insulin delivery system 14 is preferred embodiment comprised.Glucose sensor system 10 produces the sensor signal 16 of the blood sugar level 18 represented in health 20 and sensor signal 16 is supplied to controller 12.Controller 12 sensor-lodging 16 also produces order 22, and this order 22 is passed to insulin delivery system 14.Insulin delivery system 14 receives order 22 and insulin 24 is infused to health 20 by response command 22.

In general, glucose sensor system 10 comprises glucose sensor, provides electric power and produce the sensor electronics of sensor signal 16 to sensor, sensor signal 16 is passed to the sensor communication system of controller 12 and is used for the sensing system shell of electronic component and sensor communication system.

Usually, controller 12 comprises controller electronic component and to produce for the software of the order of insulin delivery system 14 and sensor-lodging 16 based on sensor signal 16 and by command routing to the controller communication system of insulin delivery system 14.

In general, insulin delivery system 14 comprises infusion apparatus and insulin 24 is infused to the infusion tube of health 20.In certain embodiments, described infusion apparatus comprises the infusion electronic component activating infusion motor according to order 22, receives the infusion communication system of the order 22 of self-controller 12 and holds the infusion apparatus shell of infusion apparatus.

In a preferred embodiment, controller 12 is installed in infusion apparatus shell, and the order 22 that infusion communication system is self-controller 12 is in the future passed to electronic circuit or the electric wire of infusion apparatus.In alternative embodiments, controller 12 is installed in sensing system shell, and sensor communication system carries electronic circuit from sensor signal 16 to the controller electronic component of sensor electronics or electric wire.In other optional embodiments, controller 12 has himself shell or it is included in utility appliance.In another optional embodiment, described controller is arranged in infusion apparatus and sensing system is all arranged in a shell.In further alternative embodiment, sensor, controller and/or infusion communication system can use cable, electric wire, fibre circuit, RF, IR or ultrasonic transmitter and receiver etc. to replace electronic circuit.

System outline

As shown in Figure 2, the preferred embodiment of the present invention comprises sensor 26, sensor stand 28, remote measurement characteristic monitor 30, sensor wire 32, infusion apparatus 34, infusion tube 36 and infusion assembly 38, and all these are worn on user's body 20.As shown in Figure 3 A and Figure 3 B, remote measurement characteristic monitor 30 comprises the monitor shell 31 of supporting printing wiring board 33, battery 35, antenna (not shown) and pickup wire cable connector (not shown).As shown in Fig. 3 D and Fig. 4, the detection end 40 of sensor 26 has the electrode 42 of exposure and this detection end 40 inserts the hypodermis 44 of user's body 20 through skin 46.Electrode 42 contacts the interstitial fluid (ISF) existed in whole hypodermis 44.As shown in figs. 3 c and 3d, sensor 26 remains on certain position by sensor stand 28, and sensor stand 28 is adhered fixed in user's skin 46.Sensor stand 28 provides the connector end 27 of sensor 26, to be connected to the first end 29 of sensor wire 32.Second end 37 of sensor wire 32 is connected to monitor shell 31.Being included in battery 35 in monitor shell 31 for the electronic component 39 on sensor 26 and printed-wiring board (PWB) 33 provides electric power.Electronic component 39 pick-up transducers signal 16 and digital sensor value (Dsig) being stored in memory, regular subsequently the digital sensor value Dsig in storer is emitted to controller 12, this controller 12 is included in infusion apparatus.

Controller 12 processes digital sensor value Dsig and produces the order 22 for infusion apparatus 34.Preferably, as shown in Figure 5, infusion apparatus 34 response command 22 is actuation plunger 48 also, and this plunger 48 forces insulin 24 from being arranged in the fluid reservoir 50 of infusion apparatus 34 inside out.In certain embodiments, the connector tip 54 of fluid reservoir 50 extends through infusion apparatus shell 52 and the first end 51 of infusion tube 36 is connected to connector tip 54.Second end 53 of infusion tube 36 is connected to infusion assembly 38.Insulin 24 is forced to by insulin pipe 36, enters infusion assembly 38 and enters health 20.As shown in Figure 6, infusion assembly 38 adheres to user's skin 46.As a part for infusion assembly 38, sleeve pipe 56 extends through skin 46 and stops in hypodermis 44, produces fully liquid and be communicated with between fluid reservoir 50 and the hypodermis 44 of user's body 20.

In alternative embodiments, closed-loop system can be a part for hospital's glucose monitoring system.No matter whether suffer from diabetes in view of before experimenter, in Intensive Care Therapy process insulinization demonstrated greatly improve wound healing, reduce bloodstream infection, renal failure and polyneuropathy mortality ratio (see, the people such as Van den Berghe G., NEJM 345:1359-67,2001, this list of references is incorporated to herein by reference), the present invention can be used for this hospital equipment to control the blood sugar level of the patient in Intensive Care Therapy.In these optional embodiments because patient in intensive care unit time (such as, ICU) vein (IV) instil and be generally implemented on the arm of patient, so closed loop glucose controls to build on existing IV and connect by backpack.Therefore, in hospital system, the IV conduit being connected directly to patient vessel's system in order to rapid delivery IV fluid also can be used for facilitating blood sampling and directly material (such as, insulin, anti-coagulants) is infused to Ink vessel transfusing space.And glucose sensor inserts by IV pipeline to provide the real-time glucose level in blood flow.Therefore, based on the type of hospital system, optional embodiment not necessarily needs described system element, such as, and the sensor 26 preferred embodiment, sensor stand 28, remote measurement characteristic monitor 30, sensor wire 32, infusion tube 36 and infusion assembly 38.On the contrary, the name submitted on September 27th, 2002 is called the provisional application the 60/414th of " Multi-lumenCatheter ", and the Standard blood glucose dosage facility described in No. 248 (its full content is incorporated to herein by reference) or blood vessel glucose sensor can be used for thinking that infusion pump controls to provide blood glucose value and existing IV connection can be used for insulin administration in patient.

Importantly, should be appreciated that the multiple combination of the equipment in hospital system together can use with closed loop controller of the present invention.Such as, described by Figure 39 B compared with the preferred system in Figure 39 A, automatic blood sugar/vein Insulin pumping system can fixed intervals (preferably, 5-20 minute) Automatic Extraction blood sample analyze the concentration of glucose of blood sample, with interval more frequently (preferably, 1 minute) extrapolation blood glucose value, and use the signal of extrapolation to calculate IV-infusion of insulin according to following controller.Automatic blood sugar/vein the Insulin pumping system of improvement can be eliminated the needs that subcutaneous sensor compensates and subcutaneous insulin compensates (as described by lead-lag compensation device hereafter).Blood Automatic Extraction and glucose assays subsequently can be completed by prior art (such as, VIA or Biostator sample blood sugar analyzer) or be completed by the system described by Figure 40.System in Figure 40 uses peristaltic pump 420 to extract blood to stride across amperometric sensor 410 (identical with the technology that sensor 26 uses), and from fluid reservoir 400, feeds back blood by increasing flushing (0.5 to 1.0ml) subsequently.Flushing can be made up of any combination of salt solution, heparin, glucose solution etc.If blood sample is to be greater than 1 minute but to be less than the interval acquiring of 20 minutes, so blood sugar detection can per minute based on extrapolate, wherein, extrapolation is carried out based on currency (n) and preceding value (n-1), to work together with the logic of the controller hereafter described in detail.For with the blood sample of the interval acquiring being greater than 20 minutes, zero-order holder (zero-order-hold) can be used for extrapolating.Based on these blood glucose values, infusion apparatus can based on the closed loop controller administration insulin hereafter more described in detail.

In other improvement of system, artificial blood sugar/vein Insulin pumping system can be used, wherein, artificial input from standard blood glucose meter (such as frequently, YSI, Beckman, etc.) blood glucose value and extrapolate this value to set up the substitution signal for calculating IV-infusion of insulin with interval (preferably, 1 minute) more frequently.Alternatively, sensor blood sugar/vein Insulin pumping system can use continuous glucose sensor (such as, blood vessel, subcutaneous, etc.), for blood sugar detection frequently.And according to controller hereinafter described, infusion of insulin can subcutaneous administration, but not the intravenous administration described in any one example in previous case.

In another optional embodiment, system element can differently divide the demand being equipped with and adapting to user with less or more equipment conbined usage of quantity and/or the function of each equipment.

Controller

Once be configured for the hardware of closed-loop system, such as above-mentioned preferred embodiment in, the effect of hardware to human body is measured by controller.In a preferred embodiment, controller 12 is designed to simulation pancreatic beta cell (beta cell).In other words, controller 12 indicates infusion apparatus 34, with following speed, insulin 24 is discharged into health 20, the concentration curve that the concentration curve that described speed causes the insulin concentration in blood to be followed producing with the fully functioning people beta cell of the blood sugar concentration responded in health 20 is similar.In further embodiment, can use " semiclosed loop " system, wherein, user needed to confirm that insulin is sent in time before any insulin of actual delivery.

The controller of simulated body to the natural insulin response of blood sugar level not only makes insulin effectively be used, but also is responsible for other body functions, because insulin has metabolism and Mitosis.Be designed to minimize glucose oscillation in body and do not consider that the controller algorithm of insulin delivering amount can cause excessive body weight increase, hypertension and arteriosclerosis.In the preferred embodiment of the present invention, controller 12 is insulin secretion pattern and regulate this pattern to change consistent with beta cell in body in analogue body.Insulin sensitivity (the S there is normal glucose tolerance (NGT), significantly changing _i) experimenter body in beta cell reaction be maintain the best insulin response of glucose homeostasis.

Beta cell and PID control

In general, in body, beta cell is characterized by " first " and " second " stage insulin response the reaction that glucose changes.As shown in fig. 23b, this two benches insulin response clearly illustrates in the hyperglycaemia clamp experiment process being applied to NGT experimenter.As shown in fig. 23 a, in hyperglycaemia clamp experiment process, glucose level is from foundation level G _bbe increased to new higher level G fast _cand be held constant at higher level G subsequently _c.The increasing degree (Δ G) of glucose affects insulin response.Four kinds of insulin response curves show the different glucose clamp level of in Figure 23 B four kinds.

The two benches insulin response of beta cell can use the proportional component analog with integration, derivative (PID) controller.Because pid algorithm is stable for various different non-medical dynamic system, so select PID controller, and pid algorithm has been found in the interference significantly changed in system dynamics and change process stable.

The insulin response of the beta cell in hyperglycaemia clamp experiment process uses each component analog beta cell of PID controller to be illustrated in Figure 24 A to Figure 24 E.The proportional component U of PID controller _pwith derivative component U _dcapable of being combined to represent first stage insulin response 440, this first stage continues a few minutes.The quadrature components U of PID controller _irepresent subordinate phase insulin response 442, this subordinate phase is stable under hyperglycaemia clamp condition increases insulin releasing.Each component is described by following formula the percentage contribution of insulin response:

Proportional component reacts: U _p=K _p(G-G _b);

Quadrature components is reacted: $U_I=K_I{\∫}_{t_0}^t(G-G_B)\mathrm{dt}+I_B,$ With

Derivative component reacts: $U_D=K_D\frac{\mathrm{dG}}{\mathrm{dt}},$

Wherein,

U _pit is the proportional component of the order being sent to insulin delivery system;

U _iit is the quadrature components of the order being sent to insulin delivery system;

U _dit is the derivative component of the order being sent to insulin delivery system;

K _pit is proportional gain factor;

K _iit is integration gain factor;

K _dit is derivative gain coefficient;

G is current blood glucose level,

G _bthe basal glucose level expected,

T is the time in past from last sensor calibration,

T ₀the time of last sensor calibration, and

I _bt ₀time basal insulin levels or can be described as U _i(t ₀).

Being combined in Figure 24 E of PID component in two stages of the insulin that simulation beta cell responds shows, as the hyperglycaemia clamp experiment of beta cell response diagram 24A.Figure 24 E shows the amplitude of first stage reaction 440 by derivative and proportional gain K _dand K _pdrive.And the amplitude of subordinate phase reaction 442 is by storage gain K _idrive.

The component of PID controller can also be expressed by its discrete form:

Proportional component reacts: ${P_{\mathrm{con}}}^n=K_P({\mathrm{SG}}_f^n-G_{\mathrm{sp}}),$

Quadrature components is reacted: ${I_{\mathrm{con}}}^n={I_{\mathrm{con}}}^{n-1}+K_I({\mathrm{SG}}_f^n-G_{\mathrm{sp}});{I_{\mathrm{con}}}^0=I_b,$

Derivative component reacts: ${D_{\mathrm{con}}}^n=K_D\mathrm{dGd}{t}_f^n,$

Wherein, K _p, K _iand K _dproportional gain factor, integration gain factor and derivative gain coefficient, SG _fand dGdt _fbe filtered sensor glucose and derivative respectively, and subscript n refer to discrete time.

Acute insulin response is necessary for preventing postprandial blood sugar from fluctuating widely.In general, insulin produces less total insulin needed for basal glucose level G/W pancake being back to expectation to the early reaction sharply increased of glucose level.This is because infusion of insulin adds by the number percent of the glucose of intake body.When concentration of glucose is higher, a large amount of insulin of infusion adds the number percent of glucose uptake, and this can make insulin effectively be used.Otherwise a large amount of insulin of infusion causes using a large amount of insulin to remove relatively a small amount of glucose when concentration of glucose is lower.In other words, the larger number percent of plurality word is greater than the larger number percent compared with decimal fractions.Less total infusion of insulin helps avoid and produce insulin resistance in user's body.And first stage insulin is considered to the early stage suppression producing hepatic glucose output.

Insulin sensitivity is not fixed and can significantly be changed in vivo, depends on the amount of exercise of health.Such as, in a research, exist in hyperglycaemia clamp experiment process, the insulin response in high training individuality (weekly the individuality of training more than 5 days) is compared with the insulin response had in the experimenter of normal glucose tolerance (NGT).As shown in fig. 25 a, the insulin response of training individuality 444 is about 1.5 times of the insulin response of NGT experimenter 446.But as shown in Figure 25 B, the glucose uptake speed of each individuality (training individual 448 or normal individual 450) is almost identical.Therefore, can infer that training individuality has twice insulin sensitivity and has the half insulin response producing identical glucose uptake individual with NGT.As shown in fig. 25 a, not only first stage insulin response 440 reduces due to motion effect, and subordinate phase insulin response 442 also demonstrates adjustment insulin sensitivity.

In a preferred embodiment, closed-loop control system can be used for insulin delivery being delivered to health to compensate insufficient beta cell played a role.Each health has the horizontal G of desirable basal plasma glucose _b.The horizontal G of desirable basal plasma glucose _band the difference between the estimated value of the horizontal G of current blood glucose is the glucose level error G that must be corrected _e.As shown in figure 26, glucose level error G _econtroller 12 is supplied to as input.

If glucose level error G _ejust (mean that the current estimated value of blood sugar level G is higher than the horizontal G of desirable basal plasma glucose _b), so controller 12 produces insulin delivery and loses one's life and make 22 to supply insulin 24 to health 20 to drive infusion apparatus 34.With regard to control loop, glucose is considered to positive, and therefore, insulin is negative.Sensor 26 detects ISF glucose level and produces sensor signal 16.Sensor signal 16 filtered and correct with the estimated value producing current blood glucose level 452.In certain embodiments, the estimated value of the horizontal G of current blood glucose adopts correcting algorithm 454 to regulate, subsequently by itself and the horizontal G of desirable basal plasma glucose _bcompare to calculate new glucose level error G _e, again start ring.

If glucose level error G _enegatively (mean that the current estimated value of blood sugar level is lower than the horizontal G of desirable basal plasma glucose _b), so controller 12 reduces or stops insulin sending, and this depends on glucose error G _equadrature components reaction whether remain positive.

If glucose level error G _ezero (mean the horizontal G of basal plasma glucose that the current estimated value of blood sugar level equals desirable _b), so controller 12 can send or can not send the order of infusion of insulin, and this depends on derivative component (whether glucose level raises or reduce) and quadrature components, and (glucose level is higher or lower than the horizontal G of basal plasma glucose _bhow long and glucose level higher or lower than the horizontal G of basal plasma glucose _bhow much).In " semiclosed loop " embodiment, user sends between infusion of insulin order at controller 12 and is pointed out.This prompting can be shown to user over the display, sounds to user, or provides system to prepare the instruction of insulin delivery to user, such as, and vibration or other senses of touch instruction.In addition, the amount of insulin to be delivered can show, other information of with or without, such as, one day inject total amount or send the potential impact to user blood glucose level by insulin.As response, user can show that insulin should be delivered or should not be delivered, such as, by select button, key or other inputs.In further embodiment, at least twice keyboard must be had to knock, and such insulin can not accidentally be sent.

In order to clearly understand the impact on control loop that health has, need to be further described in more detail the physiological action of insulin to the concentration of glucose in interstitial fluid (ISF).In a preferred embodiment, infusion apparatus 34 passes through the ISF of conduit 56 insulin delivery to the hypodermis 44 of health 20 of infusion assembly 38.Further, as shown in the block diagram of Figure 27, insulin 24 diffuses out from around the local I SF of conduit, enters blood plasma, takes a walk subsequently in major circulatory system in whole health 20.Insulin diffuses out subsequently from blood plasma, to enter between tissue also ISF, substantially throughout whole health.The membrane receptor protein of insulin 24 on systemic cell is combined and activates the membrane receptor protein on bodily tissue cell.This promotes that glucose penetrates into the cell of activation.Like this, the tissue of health 20 absorbs the glucose in ISF.When ISF glucose level reduces, glucose diffuses out from blood plasma, enters ISF to maintain concentration of glucose balance.Finally, the glucose permeation sensor film in ISF also affects sensor signal 16.

In addition, insulin has hepatic glucose output directly and indirectly affects.The insulin concentration increased reduces hepatic glucose output.Therefore, acute and direct insulin response not only contributes to the effective ingestion of glucose of health, and significantly stops liver to add glucose in blood flow.In alternative embodiments, insulin is more directly delivered to blood flow, but not interstitial fluid, such as, be delivered to vein, artery, abdominal cavity, etc.Therefore, eliminate and to move the relevant any time with insulin from interstitial fluid to blood plasma and postpone.In other optional embodiments, glucose sensor and blood or bioresorbable, and do not contact with interstitial fluid, or glucose sensor to be positioned at health outside and measure glucose by non-invasive manner.Use the embodiment of optional glucose sensor can have shorter or longer delay between blood glucose levels and the blood glucose levels of measurement.

Selection control gain

In a preferred embodiment, selection control gain K _p, K _iand K _dthe order of self-controller 12 makes infusion apparatus 34, with following speed, insulin 24 be discharged into health 20 like this, the concentration curve that the concentration curve that described speed makes the insulin concentration in blood follow to produce with the fully functioning people beta cell responding blood sugar concentration in body is similar.In a preferred embodiment, gain is selected by the insulin response that the normal glucose tolerance (NGT) observing several and have healthy normal functionality beta cell is individual.Determine that the first step of a series of controller gain is that the blood insulin concentration carrying out regular blood sugar measurement and NGT group of individuals is measured.The second, each individuality in group carries out hyperglycaemia clamp experiment, continues to carry out regular blood sugar and blood insulin concentration is measured and record simultaneously.3rd, least square curve fit is applied to the blood insulin concentration of each individuality of recorded measure of the change in time.Result is the insulin that a series of curve represents the hyperglycaemia clamp experiment of each individuality in response group.4th, described curve is for calculating the controller gain K of each individuality _p, K _iand K _d.And last, the proportional gain from each individuality together averages, thus obtain the average proportions gain K of band for controller 12 _p.Similarly, storage gain K _iwith derivative gain K _dbe averaged, thus obtain the average integral gain K being used for controller 12 _iwith average derivative gain K _d.Alternatively, other statistics values can replace mean value to use, such as, and maximal value, minimum value, high value or lower value, two or three Σ standard deviations, etc.In group, the gain of the calculating of each Different Individual can be filtered to remove exceptional data point, subsequently from statistically calculating the gain being ready to use in controller.

In instances, as shown in figs. 28 a and 28b, least square curve fit method is for generation of the insulin response curve of interior two fasted subjects of representative group.Subsequently, controller gain is calculated by the insulin response curve of two representative individual and is listed in table 1.When computing controller gain, insulin clearance be assumed that 10 (ml insulin)/minute/(kg body weight).Insulin removing speed k is the speed removing insulin in body blood flow.Finally, the measured value in the mean value use group of the gain of every type calculates, as shown in table 1.

Individual	Proportional gain, K P	Storage gain, K I	Derivative gain, K D
				a	0.000406	0.005650	0.052672
b	0.000723	0.003397	0.040403
				Mean value	0.000564	0.004523	0.046537

The PID controller gain that table 1. is calculated by the insulin response curve of two NGT individualities

Controller gain can be expressed and/or can be improved by conversion factor according to free memory of selected English or S.I. unit, floating-point or integer software simulating, software etc. by various different unit.The unit of the controller gain in table 1 arranges and is:

K _p: (mU insulin)/minute/(Kg body weight) every (mg glucose)/(dl blood plasma);

K _i: (mU insulin)/minute/(Kg body weight) every (mg glucose)/(dl blood plasma) minute; And

K _d: (mU insulin)/minute/(Kg body weight) every (mg glucose)/(dl blood plasma)/minute.

In alternative embodiments, other curve-fitting methods are used to produce insulin response curve by the measured value of blood insulin concentration.

Need the estimated value of insulin removing speed (k), whose body weight (W) and insulin sensitivity S _icontroller gain is calculated by the insulin response curve of each NGT individuality.Insulin removing speed (k) is usual and body weight is proportional and on the books in the literature.Individual insulin sensitivity S _ithe resistance to tested person of intravenous glucose can be used, hyperglycaemia clamp experiment measures, or when diabetic, individual insulin sensitivity S _iby more individual every day insulin requirements and their carbohydrates intake every day measure.

In specific embodiment, measure two parameters of each individuality, insulin sensitivity S _iwith insulin removing speed k.In other embodiments, insulin removing speed is estimated by document under the condition of given whose body weight.In other particular implementation, use longer or shorter insulin checkout time.In other embodiments, all parameters are estimated.In extra embodiment, measure one or more than a kind of parameter, estimate at least one parameter by document simultaneously.

In other optional embodiments, controller gain uses the group of individuals with similar body types to calculate.Such as, can measure that several are high, thin, the insulin of the response hyperglycaemia clamp experiment of the NGT male sex, thus the controller insulin response gain of each individuality in calculating group.Subsequently, described gain is merged statistically a series of for representational controller gain that is high, thin, the NGT male sex to produce.Can identical measurement be carried out to other groups, such as, but not limited to: short, heavy, NGT women; Medium altitude, medium body weight, the women of high degree of motion training of carrying out; 10 years old individuality of average height and weight, etc.Subsequently, their controller gain of group selection for each individual consumer is represented based on the best.In further alternative embodiment, for each individual consumer selects unique controller gain.In specific embodiment, the controller gain of user based on insulin sensitivity measured value, insulin checkout time, insulin time of occurrence, insulin concentration, body weight, body fat percentage, body metabolism or other physical traits (such as, pregnancy, age, heart condition, etc.) select.

In other optional embodiments, controller gain is estimated as user's weight W and insulin sensitivity S _ifunction.A series of observations is for proving the method.First observations is that controller gain is proportional to one another.In other words, the change that concentration of glucose is less causes less rate response U _d, less ratio response U _pless integration response U _i.Further, as shown in fig. 23b, the larger change of concentration of glucose causes proportional larger rate response U _d, proportional larger ratio U _presponse and proportional larger integration response U _i.The change of concentration of glucose affects controller response U pro rata _pIDall three components.Second observations be first stage insulin response ( ) and derivative gain K _dproportional.Further, the 3rd observations is that two constants can easily obtain from the information published document or xsect by population is measured.Two constants are disposal indexes (DI) of insulin removing speed (k) of the people of given body weight and the people of given concentration of glucose change.

When there is the information source needed for multiple calculating insulin removing speed k, a source is that the article " Insulin clearance during hypoglycemia in patientswith insulin-dependent diabetes mellitus " that Kollind M writes (is published in Horm Metab Res, 1991July; 23 (7): 333-5).Insulin removing speed k is obtained divided by steady state blood plasma insulin concentration by the insulin of infusion.Insulin independent of whose body weight removes constant A _kobtain divided by whose body weight by insulin removing speed k (obtaining from particular individual measurement).Insulin removes constant A _kusually be identical for all mankind, except making situation about losing weight, such as, in individual infected by HIV, the disease of other influences metabolism, etc. after.

The article " Quantification of the relationship between insulin sensitivity andbeta-cell function in human subjects.Evidence for a hyperbolic function " that the disposal index (DI) of the people of given concentration of glucose change can be write from people such as Khan S E (is published in Diabetes, 1993 November; 42 (11): 1663-72) information acquisition of display in.

Dispose both index D I and insulin removing speed k by testing direct measurement.Dispose index D I to measure under the condition of given first stage insulin response and individual insulin sensitivity, described first stage insulin response is measured by glucose clamp experiment, and described individual insulin sensitivity is by insulin sensitivity thermometrically.Insulin removing speed k removes thermometrically by insulin.Glucose clamp test and insulin removing test describe and are well known in the art in above-mentioned article.Insulin sensitivity S _ithe resistance to tested person of intravenous glucose or hyperglycaemia clamp thermometrically can be used.

In view of these observationss, subsequently, following parameters can be calculated by the insulin response of the NGT individuality to glucose clamp: desirable first stage insulin response k _dwith K _pratio and the ratio of KD and KI.Subsequently, derivative gain K _dby using constant k and DI by first stage insulin response calculate.Further, last K _pand K _ik can be used _dwith K _pratio and K _dwith K _iratio calculate.

First stage insulin response can observe in NGT individuality, as the insulin response area under a curve in about first process of 10 minutes of glucose clamp experiment.The increase of the concentration of glucose in glucose clamp experiment process is Δ G=(G-GB), and wherein, G is equivalent to Gc, the concentration of glucose in clamp experiment process, G _bbasal glucose concentration before being clamp experiment.

First stage insulin response importance emphasized by following research, described research shows, in the experimenter with normal glucose tolerance (NGT), first stage insulin response with insulin sensitivity (S _i) product be called to dispose the constant of index, therefore,

$\mathrm{\φ}1=\frac{\mathrm{DI}}{S_I}.$

For different Δ G, have different produce different DI thus.But, ratio DI/ Δ G substantially constant, even for there is the Different Individual of different insulin sensitivity, ratio DI/ Δ G also substantially constant.

Insulin sensitivity S _ibe defined as the number percent of the concentration of glucose that the bodily tissue for the insulin of specified rate will absorb.Beta cell is by regulating it in first stage insulin response the amount of the insulin secreted in process and the change of natural adaptation insulin sensitivity.This illustrates that health seeks best glucose-tolerant level naturally.The natural insulin response of the controller simulated body more exactly of this feature of simulation beta cell.

Instantaneous insulin response (RI) can in insulin removing speed (k) and first stage insulin response calculate under given condition,

Insulin removing speed k and body weight (W) proportional, therefore, use proportionality constant A _ksubstitute k with user's weight W and use DI and S _iratio replace obtain following formula:

Instantaneous insulin response R _ialso derivative gain K can be expressed as _dwith the product of concentration of glucose changes delta G, R _i=K _dΔ G.

Set in two formula, R _ibe equal to each other, and solve K _dobtain:

As mentioned above, DI/ Δ G and A _kcan by the data acquisition in the document published or calculating

$Q=\frac{A_k\mathrm{DI}}{\mathrm{\ΔG}}$

$K_D=\frac{W}{S_I}Q$ Constant.Merging constant is single constant Q, obtain derivative gain K _dformula, K _duser's weight W and user's insulin sensitivity S _ifunction.

Once calculate derivative gain K _d, so, proportional gain and storage gain are calculated by usage ratio.K _d/ K _pratio can be set as the main time constant of insulin action, the time is from 10 minutes to 60 minutes, but more common, the time from 20 minutes to 40 minutes, and preferably 30 minutes.Such as, the time constant of 30 minutes is used, at given K _dcondition under calculating K _p, obtain following relationship:

in a similar fashion, K _d/ K _iratio can be set as the average proportions measured by NGT groups of individuals.And K _iby K _dcalculate.

In specific embodiment, user is by their body weight W and insulin sensitivity S _iinput to and comprise in the equipment of controller.Subsequently, controller gain is automatically calculated out and is used by controller.In alternative embodiments, individual by user's weight W and insulin sensitivity S _iinput to equipment and described equipment provides the information of calculated gains to controller.

The present invention carries out studying confirming using glucose sensor can the insulin response of rendering individual as input.Under study for action, when carrying out hyperglycaemia clamp experiment to NGT individuality, obtain glucose and insulin measured value.The measured value of the glucose level shown in Figure 29 A is used as the input of the mathematical model of the analog pid insulin response controller set up.As shown in fig. 29b, the insulin dose indicated by controller of response glucose clamp is very exactly close to the actual insulin apparent value in NGT individuality.The insulin concentration calculated by the regular blood sample 456 taking from individuality in test process represents in Figure 29 B with round dot.The output carrying out the mathematical model of the simulation insulin response of self-controller instruction is presented in Figure 29 B with solid line 458.

In research process, three kinds of different equipment are for measuring individual blood sugar.Represent in Figure 29 A with round dot from the blood glucose meter reading 460 taking from individual regular blood sample.Two kinds of MiniMed sensors (such as, hereinafter those described in the part of " sensor " of title) are placed in individual hypodermis, and sensor reading 462,464 is presented in Figure 29 A with solid line.Sensor reading 462,464 postpones slightly relative to blood glucose meter reading 460.This delay be most possibly caused by the delay between blood sugar and interstitial fluid (ISF) glucose and by use wave filter (if necessary) basic correction.In this study, do not postponed by filter correction, and postpone the ability having no significant effect controller instruction insulin response, described insulin response is mated in the natural reaction of NGT individuality.This research display PID insulin response controller model is the good least model of insulin secretion, and the two benches response of healthy beta cell caught by this model.Forseeable, the correction of delay only increases the degree of accuracy of model.

The fuzzy logic selected is carried out between the gain of many group controllers

In a preferred embodiment, a group controller gain is used for particular individual.In alternative embodiments, use the controller gain more than a group, and fuzzy logic is used for carrying out selecting between the gain of many group controllers and determining when from a group controller gain transition to another group.In specific Alternate embodiments, if glucose level is higher or lower than desirable glucose foundation level, so, controller gain is different.In other optional embodiments, if glucose level improves or reduces, so controller gain is different.The reason of different group gain comes from Physiologic Studies, and this research shows that the closedown of beta cell is opened faster than it.In other optional embodiments, whether controller gain is different higher or lower than whether desirable glucose foundation level and glucose level improve or reduce based on glucose level, and this produces four group controller gains.In other optional embodiments, the amplitude that controller gain drifts about based on hypoglycemia changes.In other words, the gain of the larger change for glucose is different from for the controller gain of the less change of glucose.

Self-tuning controller gain

Further embodiment can comprise one or more than one gain of self-regulation K _p, K _i, K _dto adapt to the controller of insulin sensitivity change.In specific embodiment, by the baseline measurement of glucose level and the horizontal G of desirable basal glucose _bcompare.Such as, from previous glucose level measured value, the horizontal G of desirable basal glucose is deducted _b.Subsequently, suing for peace to any negative value in schedule time window (is in fact merge lower than the horizontal G of basal glucose _bglucose level measured value).If the summation obtained is greater than the hypoglycemia integral threshold selected in advance, so controller gain adds the factor (1+ α).On the contrary, if measure in schedule time window higher than the horizontal G of basal glucose _bthe integration of glucose level measured value be greater than the hypoglycemia integral threshold selected in advance, so controller gain reduces the factor (1-α).

In specific embodiment, by the schedule time window of its estimation concentration of glucose integration normally 24 hours, and controller gain carried out regulating (if necessary) at the end of each schedule time window.In alternative embodiments, the integration of glucose level measured value, by moving time-window Continuous plus, and if integration exceeds threshold value, regulates gain so immediately.In specific embodiment, moving time-window is 1 hour, and time window can restart at any time when gain is conditioned.In other optional embodiments, based on sensor degree of accuracy, the pace of change of individual insulin sensitivity, the computing power of hardware, etc., time window is longer or shorter.

In certain embodiments, regulated quantity (α) is 0.01.In alternative embodiments, regulated quantity α is based on sensor degree of accuracy, individual insulin sensitivity pace of change, transducer sensitivity S _ipace of change, etc. greater or lesser.In other optional embodiments, the quantitative change that regulated quantity α exceedes threshold value based on the integration of glucose level measured obtains greater or lesser.Like this, if the glucose level G measured significantly departs from desirable blood sugar level G _b, so regulate gain with larger amount, if the glucose level G measured is more close to desirable blood sugar level G _b, so regulate gain with less amount.In the embodiment of Additional optional, controller uses Kalman (Kalman) wave filter.

State change feedback

Although determine that the primary signal of the insulin response of beta cell is glucose, also there is the presumption effect of the insulin suppressing insulin secretion itself.This effect can be directly related with the insulin concentration (IP (t)) in blood plasma or mediated by the signal (IEFF (t)) that some and insulin action are proportional.Beta cell may these signals of direct-detection (that is, direct-detection insulin concentration and the second signal proportional with insulin action (such as free fatty acid)).Feedback sort from these M signals is similar to the signal being called state change feedback, namely be exactly following feedback: by means of this feedback, the feedback of each M signal of the variable (being glucose in this case) be controlled and variation (insulin concentration in blood plasma and interstitial fluid) together uses.By such feedback, can make undesirable slower dynamic process seem than they self fast.Such as, if beta cell insulin secretion is subject to the suppression of the signal proportional with the insulin concentration in interstitial fluid (described beta cell insulin secretion plays a role) wherein, the delay between blood plasma and interstitial fluid so can be made to seem shorter.For artificial closed loop algorithm, or for " semiclosed loop " algorithm, this beneficial effect realizes by using " watchers " (known in the past insulin send history and estimate the mathematical formulae of the insulin concentration in health various piece).In " semiclosed loop " algorithm, algorithm is identical with closed loop algorithm, but needs user to confirm before any insulin of actual administration.By using state change feedback, the onset faster than actual insulin of the insulin in insulin pump may be made.

In order to estimate subcutaneous insulin concentration, plasma insulin concentrations and insulin action, can use following formula:

$\frac{{\mathrm{dI}}_{\mathrm{SC}}}{\mathrm{dt}}={\mathrm{\α}}_1(I_D-I_{\mathrm{SC}})$

$\frac{{\mathrm{dI}}_P}{\mathrm{dt}}={\mathrm{\α}}_2(I_{\mathrm{SC}}-I_P)$

$\frac{{\mathrm{dI}}_{\mathrm{EF}}}{\mathrm{dt}}={\mathrm{\α}}_3(I_P-I_{\mathrm{EF}})$

Wherein, I _sCthe estimated value of the insulin concentration of subcutaneous space Plays, I _pthe estimated value of the insulin concentration of blood plasma Plays, I _eFthe estimated value to the insulin that glucose works, α ₁the velocity constant that insulin sends between subcutaneous insulin chamber, α ₂the velocity constant between subcutaneous insulin and blood plasma chamber, α ₃it is the velocity constant between blood plasma chamber and insulin action.I _dbe the insulin sent, it can be three state change (I _sC, I _pand I _eF) function.

In specific embodiment, according to following formula, what the basal rate that open loop is fixed added that user requires inject can cause injecting increases a certain amount of and basal rate reduces identical amount subsequently:

${I_D}^{\′}=(1+{\mathrm{\γ}}_1+{\mathrm{\γ}}_2+{\mathrm{\γ}}_3)I_D-{\mathrm{\γ}}_1{I}_{\mathrm{SC}}{-\mathrm{\γ}}_2{I}_P-{\mathrm{\γ}}_3{I}_{\mathrm{EF}}$

Wherein, I _dthat basis (U/h) that user requires adds and injects curve (U) and I _dit is the curve of state feedback regulation.Note that for the drift of given dynamics, (the I of requirement _darea under curve) insulin total amount and send (I _darea under curve) insulin total amount identical.At this, γ ₁, γ ₂and γ ₃feedback of status gain (scalar).These gains are carefully selected, pump corrects its delivery rate and diffuses into patient hypodermic layer relevant delay with insulin by injecting to compensate, pump corrects its delivery rate and diffuses to the relevant delay of blood plasma to compensate with insulin bolus, and pump corrects its delivery rate diffuses to from injecting the impact/act on relevant delay that health produces reality with insulin to compensate.Therefore, by estimating the amount or actual glucose level (the state variable I to patient of estimation that inject the insulin entered in hypodermic layer, in blood plasma _sC, I _pand I _eF) amount of insulin that works, can optimize in time insulin delivery to patient.Using state is fed back, and injects and adds (1+ γ ₁+ γ ₂+ γ ₃) amount, this amount removes (-γ gradually from insulin is in the future sent ₁i _sC-γ ₂i _p-γ ₃i _eF).Therefore, the Insulin Pharmacokinetics curve demonstrated seems faster.This insulin type with the very fast onset of exploitation with, but it sends distribution by the insulin rearranged per unit and inject, by sending amount and remove the additional quantity in later stage and algorithmically realize in more early stages.Three kinds of gains can be selected as traveling time and postpone (1/ α ₁, 1/ α ₂with 1/ α ₃) to any optional position.In control theory, this is called as POLE PLACEMENT USING.

Feedback of status can be used for Open loop and closed loop insulin and to send in algorithm and to send algorithm for " semiclosed loop ".Feedback of status can with the closed loop controller coupling of proportional, integral-derivative (PID) or any other type.γ ₁be and I _sGthe feedback gain be multiplied, γ ₂be and I _pthe feedback gain be multiplied, and γ 3 is and I _eFthe feedback gain be multiplied.

The physical state space form directly obtained from above-mentioned formula is:

$\left\{\begin{array}{c}\left[\begin{array}{c}{\stackrel{\·}{I}}_{\mathrm{SC}}\\ {\stackrel{\·}{I}}_P\\ {\stackrel{\·}{I}}_{\mathrm{EF}}\end{array}\right]=\left[\begin{array}{ccc}-{\mathrm{\α}}_1& 0& 0\\ {\mathrm{\α}}_2& -{\mathrm{\α}}_2& 0\\ 0& {\mathrm{\α}}_3& -{\mathrm{\α}}_3\end{array}\right]\·\left[\begin{array}{c}{I}_{\mathrm{SC}}\\ I_P\\ I_{\mathrm{EF}}\end{array}\right]+\left[\begin{array}{c}{\mathrm{\α}}_1\\ 0\\ 0\end{array}\right]\·I_D\\ I_D=\left[\begin{array}{ccc}0& 0& 0\end{array}\right]\·\left[\begin{array}{c}{I}_{\mathrm{SC}}\\ I_P\\ I_{\mathrm{EF}}\end{array}\right]+\left[\begin{array}{c}1\\ 0\\ 0\end{array}\right]\·I_D\end{array}\right.\mathrm{or}\left\{\begin{array}{c}\stackrel{\·}{x}=\mathrm{Ax}+\mathrm{Bu}\\ y=\mathrm{Cx}+\mathrm{du}\end{array}\right.$

Finite difference form calculates as follows (wherein, e ^xrepresent exponential function):

Definition: $k_1=e^{-{\mathrm{\α}}_{1}T},k_2=e^{-{\mathrm{\α}}_{2}T},k_3=e^{-{\mathrm{\α}}_{1}T}$

I _sC(i)=(1-k ₁) (I _d(i-1))+k ₁i _sC(i-1) (formula 1b)

I _p(i)=(1-k ₂) (I _sC(i))+k ₂i _p(i-1) (formula 2b)

I _eF(i)=(1-k ₃) (I _p(i))+k ₃i _eF(i-1) (formula 3b)

Laplce's table (Laplace Form) is as follows, and wherein, s represents and uses in Laplace formula determinant:

$\frac{I_{\mathrm{SC}}}{I_D}=\frac{{\mathrm{\α}}_1}{s+{\mathrm{\α}}_1}$ (formula 1c)

$\frac{I_P}{I_{\mathrm{SC}}}=\frac{{\mathrm{\α}}_2}{s+{\mathrm{\α}}_2}$ (formula 2c)

$\frac{I_{\mathrm{EFF}}}{I_P}=\frac{{\mathrm{\α}}_3}{s+{\mathrm{\α}}_3}$ (formula 3c)

$\frac{I_P}{I_D}=\frac{{\mathrm{\α}}_1{\mathrm{\α}}_2}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)}$ (formula 4)

$\frac{I_{\mathrm{EFF}}}{I_D}=\frac{{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)}$ (formula 5)

In order to obtain transformation by feedback of status, contrast formula is as follows, and wherein E represents the error (G-G between actual glucose concentration and desirable concentration of glucose _d):

I _d=PIDE-γ ₁i _sC-γ ₂i _p-γ ₃i _eFF(formula 6)

In (formula 6), replace formula (formula 1c), (formula 4) and (formula 5) also arranges again, and obtain following transfer function, wherein, GM is gain multiplier:

$\frac{I_D}{E}=\frac{\left(\mathrm{GM}\right)\left(\mathrm{PID}\right)(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\γ}}_1(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\γ}}_2(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3{\mathrm{\γ}}_3}$ (formula 7)

$\frac{I_{\mathrm{SC}}}{E}=\frac{\left(\mathrm{GM}\right)\left(\mathrm{PID}\right){\mathrm{\α}}_1(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\γ}}_1(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\γ}}_2(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3{\mathrm{\γ}}_3}$ (formula 8)

$\frac{I_P}{E}=\frac{\left(\mathrm{GM}\right)\left(\mathrm{PID}\right){\mathrm{\α}}_1{\mathrm{\α}}_2(s+{\mathrm{\α}}_3)}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\γ}}_1(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\γ}}_2(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3{\mathrm{\γ}}_3}$ (formula 9)

$\frac{I_{\mathrm{EFF}}}{E}=\frac{\left(\mathrm{GM}\right)\left(\mathrm{PID}\right){\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3}{(s+{\mathrm{\α}}_1)(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\γ}}_1(s+{\mathrm{\α}}_2)(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\γ}}_2(s+{\mathrm{\α}}_3)+{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3{\mathrm{\γ}}_3}$ (formula 10)

The calculating of gain multiplier also obtains with state change feedback method.When using state change feedback, gain multiplier (GM) is scalar, and it forces step-by-step movement to respond to reach identical stationary value, no matter whether using state is fed back.In other words, GM guarantees that total specified rate that per unit is injected is identical in both cases.In feedback of status situation, more insulin gave in early stage, but this extra insulin removed in the later stage.In order to calculate the GM in particular implementation, use " the end value rule " from control system.Described end value rule has set forth the stable state of any transfer function T (s) in order to estimate given any input X (s), and the stable state of response input is exported and provided by following formula:

y _SS(t→∞)＝lim _s→0(sT(s)X(s))

Laplce's table of step-by-step movement input passes through provide, and the stable state scheme of end value rule is reduced to:

y _SS(t→∞)＝lim _s→0(T(s))

When there is no feedback of status, (γ ₁, γ ₂and γ ₃=0), stable state scheme can obtain from formula 7, and it is as follows:

I _d(t → ∞)=1 (formula 11)

Under the condition of feedback of status not having gain correction factor, stable state scheme is:

$I_D(t\→\∞)=\frac{1}{1+{\mathrm{\γ}}_1+{\mathrm{\γ}}_2+{\mathrm{\γ}}_3}$ (formula 12)

GM is estimated as the ratio of formula 12 and formula 11 subsequently, thus obtains: GM=1+ γ ₁+ γ ₂+ γ ₃.

Using state change feedback, determines the closed-loop control formula that limit is arranged and feedback of status gain.Specifically, formulae discovery gain is sent by above-mentioned insulin.In specific embodiment, they are determined as follows: first, and by feedback of status, the denominator of formula 7, formula 8, formula 9 and formula 10 is:

D＝s ³+(α ₁+α ₂+α ₃+γ ₁α ₁)s ²+

(α ₁α ₂+(α ₁+α ₂)α ₃+γ ₂α ₁α ₂+(α ₂+α ₃)γ ₁α ₁)s+

(α ₁α ₂α ₃+ γ ₃α ₁α ₂α ₃+ γ ₂α ₁α ₂α ₃+ γ ₁α ₁α ₂α ₃) (formula 14)

In order to obtain the system pole in formula 7, formula 8, formula 9 or formula 10, D can be set to equal 0, thus obtains characteristic formula:

s ³+(α ₁+α ₂+α ₃+γ ₁α ₁)s ²+

(α ₁α ₂+(α ₁+α ₂)α ₃+γ ₂α ₁α ₂+(α ₂+α ₃)γ ₁α ₁)s+

(α ₁α ₂α ₃+ γ ₃α ₁α ₂α ₃+ γ ₂α ₁α ₂α ₃+ γ ₁α ₁α ₂α ₃)=0 (formula 16)

If the square root of ideal system limit or formula 16 is by eigenvalue λ ₁, λ ₂and λ ₃define, so characteristic formula can be written as:

(s-λ ₁)(s-λ ₂)(s-λ ₃)＝0

Increase and collect the similar power of s, formula 16 can be written as:

S ³-(λ ₁+ λ ₂+ λ ₃) s ²+ (λ ₁λ ₂+ λ ₁λ ₃+ λ ₂λ ₃) s-λ ₁λ ₂λ ₃=0 (formula 17)

Be set as being equal to each other by the coefficient of the similar power of s, we obtain formula system:

α ₁+ α ₂+ α ₃+ γ ₁α ₁=-(λ ₁+ λ ₂+ λ ₃) (formula 18)

α ₁α ₂+ α ₁α ₃+ α ₂α ₃+ γ ₂α ₁α ₂+ γ ₁α ₁(α ₂+ α ₃)=λ ₁λ ₂+ λ ₁λ ₃+ λ ₂λ ₃(formula 19)

α ₁α ₂α ₃+ γ ₃α ₁α ₂α ₃+ γ ₂α ₁α ₂α ₃+ γ ₁α ₁α ₂α ₃=λ ₁λ ₂λ ₃(formula 20)

This produces three formula and three unknown number γ ₁, γ ₂and γ ₃.Therefore, unknown gain can according to desirable limit λ ₁, λ ₂, λ ₃, and system time constant α ₁, α ₂and α ₃answered.When insulin appears in different chamber, these formula can make us control the desirable pharmacokinetic of insulin:

${\mathrm{\γ}}_1=\frac{-({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3+{\mathrm{\α}}_1+{\mathrm{\α}}_2+{\mathrm{\α}}_3)}{{\mathrm{\α}}_1}$

${\mathrm{\γ}}_2=\frac{{\mathrm{\λ}}_1{\mathrm{\λ}}_2+{\mathrm{\λ}}_1{\mathrm{\λ}}_3+{\mathrm{\λ}}_2{\mathrm{\λ}}_3-{\mathrm{\α}}_1{\mathrm{\α}}_2-{\mathrm{\α}}_1{\mathrm{\α}}_3-+{\mathrm{\α}}_2{\mathrm{\α}}_3({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3+{\mathrm{\α}}_1+{\mathrm{\α}}_2+{\mathrm{\α}}_3)({\mathrm{\α}}_2+{\mathrm{\α}}_3)}{{\mathrm{\α}}_1{\mathrm{\α}}_2}$

$\begin{array}{c}{\mathrm{\γ}}_3=\frac{-{\mathrm{\λ}}_1{\mathrm{\λ}}_2{\mathrm{\λ}}_3}{{\mathrm{\α}}_1{\mathrm{\α}}_2{\mathrm{\α}}_3}-\frac{{\mathrm{\λ}}_1{\mathrm{\λ}}_2+{\mathrm{\λ}}_1{\mathrm{\λ}}_3+{\mathrm{\λ}}_2{\mathrm{\λ}}_3-{\mathrm{\α}}_1{\mathrm{\α}}_2-{\mathrm{\α}}_1{\mathrm{\α}}_3-{\mathrm{\α}}_2{\mathrm{\α}}_3+({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3+{\mathrm{\α}}_1+{\mathrm{\α}}_2+{\mathrm{\α}}_3)({\mathrm{\α}}_2+{\mathrm{\α}}_3)}{{\mathrm{\α}}_1{\mathrm{\α}}_2}\\ \frac{({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3+{\mathrm{\α}}_1+{\mathrm{\α}}_2+{\mathrm{\α}}_3)}{{\mathrm{\α}}_1}-1\end{array}$

Therefore, by above-mentioned calculating, can gain be calculated and gain can be used in the dominated formulate that insulin sends:

I _D＝PID·E-γ ₁I _SC-γ ₂I _P-γ ₃I _EF

PID is the output of the PID controller of any other closed loop (or " semiclosed loop ") controller.Gain calculates once generally, but can calculate more continually, if desired.Dominated formulate can carry out calculating or Continuous plus after predetermined amount of time on the basis of repeating.Such as, but not limited to, calculating once for every five minutes, every 30 minutes or every 60 minutes.Only has state changing unit (γ ₁i _sC-γ ₂i _p-γ ₃i _eF) can carry out upgrading or renewable whole formula.By upgrading dominated formulate, can continuative improvement insulin sending to patient.

The control feedback block diagram of the embodiment of the pump of using state change feedback shows in Figure 42.As shown in the figure, by the ideal glucose G of patient _d600 input PID controller 610.The output of PID controller is that insulin sends value I _d601.As mentioned above, except insulin sends value, frame also calculates subsequently should to inject actual delivery how much insulin to patient and should remove how many from basal rate.Each Discrete time intervals point Ti (T1 620, T2 630 and T3 640), calculate and enter the amount of the insulin of hypodermic layer to provide I from pump _sC620.This value is multiplied by γ ₁605 (or by γ ₁605 decompose) and this value is deducted from the output of PID controller, thus provide the desirable insulin levels of improvement based on subcutaneous insulin concentration (by other formula following).At each Discrete time intervals point Ti, calculate and enter the amount of the insulin in blood plasma to provide I by subcutaneous chamber _p603.This value is multiplied by γ ₂606 (or by γ ₂606 decompose) and this value is deducted from the output of PID controller, thus determine the desirable insulin levels of improvement based on plasma insulin concentrations.At each Discrete time intervals point Ti, calculate the amount of the amount of the actual insulin played a role or the effective insulin chamber from the insulin in blood plasma, thus I is provided _eF604.This value is multiplied by γ ₃607 (or by γ ₃607 decompose) and this value is deducted from the output of PID controller, thus determine the desirable insulin levels of improvement based on effective insulin.Actual delivery to the insulin of experimenter 650 can change the blood sugar G of user 608 subsequently, and this is measured by sensor 660 subsequently and compares with ideal glucose 600.

Figure 43 to Figure 46 shows the action diagram of feedback of status.Figure 43 shows the impact on basal insulin delivery rate using above-mentioned algorithm to realize.Inject at time point 0.Lines 700 represent that insulin when not having using state to feed back is sent.These lines are identical with the conventional delivery of insulin bolus and be shown as 0.0000, because it does not change the amount of the basal rate sent.Other three lines represent when all feedback of status are positioned over gain gamma ₁, γ ₂or γ ₃in one middle time insulin delivery rate change in time.As can be seen from Figure, if all feedback of status are positioned over gain gamma ₁in (for hypodermic layer), basal insulin delivery rate 701 (with standard base speed about) start lower and move to zero boundary gradually or there is no the speed of feedback of state, because reach steady state (SS).If all feedback of status are positioned over gain gamma ₂in (for plasma layer), basal insulin delivery rate 702 is started from scratch, and drops to lower and gos up gradually to zero boundary subsequently, because reach steady state (SS).If all feedback of status are positioned over gain gamma ₃in (for insulin action/effect), basal insulin delivery rate 703 is started from scratch, and drops to lower, but than all γ ₂delivery rate is lower, and gos up gradually to zero boundary subsequently, because reach steady state (SS).In all cases, insulin always send identical.

Figure 44 shows feedback of status that per unit injects to the effect of subcutaneous insulin.In other words, zero time point to patient give insulin bolus and the amount that drawings show insulin in hypodermic layer from the speed of injecting to being reduced to zero.Lines 705 display does not have the amount of the insulin under the condition of feedback of status in hypodermic layer over time.Lines 706 show all feedback of status and are positioned over gain gamma ₁time middle, in hypodermic layer, the amount of insulin is over time.Lines 707 show all feedback of status and are positioned over gain gamma ₂the amount of insulin time middle in hypodermic layer over time.Lines 708 show all feedback of status and are positioned over gain gamma ₃the amount of insulin time middle in hypodermic layer over time.

Figure 45 shows feedback of status that per unit injects to the effect of plasma insulin.In other words, zero time point give insulin bolus and the amount that drawings show the insulin in plasma layer from injecting, increase (when moving to plasma layer from insulin injection to insulin from hypodermic layer, producing delay slightly) from zero, reach the speed that its peak value also gets back to zero subsequently.Lines 710 show the amount of the insulin do not had under feedback of status condition in blood plasma over time.Lines 711 show all feedback of status and are positioned over gain gamma ₁the amount of insulin time middle in blood plasma over time.Lines 712 show all feedback of status and are positioned over gain gamma ₂the amount of insulin time middle in blood plasma over time.Lines 713 show all feedback of status and are positioned over γ ₃the amount of insulin time middle in blood plasma over time.

Figure 46 shows feedback of status that per unit injects to the impact of insulin action.In other words, give insulin bolus to patient and drawings show the amount of injecting health produces the insulin of insulin action to start from scratch (enter hypodermic layer at insulin injection and play through blood plasma in the process of insulin action and be delayed) at zero time point, rise to its maximum point and be reduced to zero speed.Under lines 715 show the condition not having feedback of status, insulin action over time.Lines 716 show all feedback of status and are positioned over gain gamma ₁time middle, insulin action over time.Lines 717 show all feedback of status and are positioned over gain gamma ₂time middle, insulin action over time.Lines 718 show all feedback of status and are positioned over gain gamma ₃time middle, insulin action over time.

Figure 47 and Figure 48 compares the insulin state change feedback of conbined usage PID closed loop controller and is used alone PID closed loop controller (not having insulin state to change feedback).Figure 47 shows patient glucose's concentration of simulation over time.8 hours, 13 hours, 18 hours, 22 hours and 32 hours dining.Use with the concentration of glucose of the PID of insulin feedback of status as shown in lines 800.Use without the concentration of glucose of the PID of insulin feedback of status as shown in lines 801.With regard to concentration of glucose, preferably, do not allow the concentration of glucose of patient too high or too low, therefore, the closed loop program of high value and lower value more can be avoided better.As shown in figure 47, pass in time, the concentration of glucose with the PID of insulin feedback of status is used to improve in time (PID relative to using not with insulin feedback of status), because, it is less that concentration of glucose passes change in time, and making patient have more stable glucose level will reduce hyperglycaemia and hypoglycemic event greatly.Figure 48 shows the averaging analog insulin delivery curves from the system identical with the system of Figure 47.Lines 810 represent to use and send with the insulin of the PID of insulin feedback of status.Lines 811 represent and use the insulin of the PID not with insulin feedback of status to send.As can be seen from the figure, use the insulin with the PID of insulin feedback of status to send and comprise more spiking and dropping signal, this is produced by feedback of status.

Betterment PID controller is to comprise integrator seepage

In a preferred embodiment, PID controls response by constant-gain component K _p, K _i, K _ddescribe.Control response although preferred and ensure that (that is, Steady state glucose deducts desirable basal glucose (G to zero steady-state error _b) equal 0), but quadrature components makes FEEDBACK CONTROL go to stablize, because interim slow reduction does not occur insulin response, and quadrature components simulates the increase of insulin response.Under the condition of not carrying out any correction, quadrature components has excessively estimates the trend that insulin response increases.Because Steady state glucose and G _bbetween less difference usually insulin response control in be acceptable, so, integrator device seepage can be merged to reduce the amplitude of destabilization to the optional modeling of quadrature components.Specifically, U _it the change of () is by the item with glucose error-proportional with at U _iamplitude ratio in the item of seepage describe.This can express in following formula:

$\frac{{\mathrm{dU}}_I}{\mathrm{dt}}{K}_I(G-G_B)-K_{\mathrm{LEAK}}{U}_I$

Wherein, initial conditions are U _i(t0).

Parameter K _lEAKthe zero count constant (τ of leak rate _lEAKminute=1/K _lEAK), wherein, τ _lEAKbe can based on empirical data setting regulating parameter, and can with other gain component K _p, K _i, K _dassociation.But, present stage artificial beta cell τ _lEAKthat user inputs.U _ialso express with discrete form by standard method.

Controller rearmounted (lead-lag) compensator

In a preferred embodiment, order and send from controller, and though insulin delivery system by infusion of insulin to which position of health.In fact, suppose that insulin is directly sent and enter blood flow, used immediately by health, or insulin is delivered to certain position of health but not any time that blood flow causes postpones by regulating K _p, K _iand K _dcompensate.In this case, order is simulation beta cell insulin secretion curve usually, and it is the example shown in Figure 35 A.Further, because beta cell excreting insulin directly enters blood flow, so beta cell insulin secretion curve is desirable plasma insulin concentrations curve.But insulin delivery latency can make desirable plasma insulin concentrations curve deformation, as shown in Figure 35 B.Insulin delivery latency order is given the time quantum between time that insulin delivery system arrives blood plasma with the moment of infusion of insulin and insulin.Insulin delivery latency can be caused by diffusion delays, is represented by the circle 528 with arrow in Figure 20, and shown delay is the time that insulin that infusion enters tissue diffuses into needed for blood flow.Other principal elements of insulin delivery latency can comprise: after receiving the order of infusion of insulin, delivery system insulin delivery is to the time of health, insulin is once the time entering blood flow and just spread in the whole circulation system, and/or the delay that other machineries or physiological mechanism cause.In addition, health remove insulin, even when from insulin delivery system in body during insulin delivery dosage health also remove insulin.Because insulin is removed constantly by health from blood plasma, so the insulin dose at least part of (if not a large amount of) crossing insulin dose or the delay being slowly delivered to blood plasma was eliminated before whole insulin dose arrives blood plasma completely.Therefore, the insulin concentration curve in blood plasma never reaches identical peak value (also not following identical curve), if do not postponed, the insulin concentration curve in blood plasma can reach peak value.Under insulin dose all once being sent at zero time point the condition entering blood plasma, the insulin concentration in blood plasma almost raises (not shown) instantaneously and reduces in time with exponential form, according to formula because health removes insulin subsequently as shown in figure 36, wherein:

C _pthe insulin concentration in blood plasma,

I ₀the quality being directly delivered to the insulin dose of blood plasma at zero time point,

V _pthe Plasma volumes in health,

P ₁the zero count constant that insulin is removed, and

T directly sends from insulin dose the time entered blood plasma in the past.

The time constant P1 that insulin is removed can use following formulae discovery: wherein,

K is insulin removing speed, and

V _pit is the volume of blood plasma in health.

Or, the time constant P that insulin is removed ₁the individual of own insulin is not produced and insulin concentration in the blood sample of this individuality of periodic measurement subsequently obtains by being supplied to by insulin.Subsequently, use exponential curve fitting formula to produce the mathematic(al) representation of the optimum fit curve of insulin concentration measured value, and observe the time constant in described mathematic(al) representation.

As shown in figure 36b, give hypodermis by identical insulin dose (once sending all insulin doses at zero time point), and directly do not send and enter blood plasma, the insulin concentration in blood plasma is along with insulin is from interstitial fluid I _sFdiffuse into blood plasma and start slow rising.While insulin enters blood plasma, health removes insulin from blood.When the speed that insulin enters blood plasma has exceeded insulin removing speed, the insulin concentration in blood plasma has continued to increase.When insulin removing speed has exceeded insulin from interstitial fluid I _sFwhen entering the speed of blood plasma, the insulin concentration in blood plasma starts to reduce.Therefore, insulin delivery is fed into interstitial fluid and non-immediate to send the result entering blood flow be that insulin concentration in blood plasma spreads in time, but not to postpone after almost reaching peak value instantaneously.

Given be delivered to the condition of hypodermic insulin dose under, two exponential formula can be used for the insulin concentration in simulating blood plasma:

$C_P=\frac{I_{0}D}{V_p{V}_{\mathrm{ISF}}(P_3-P_2)}(e^{-P_{2}t}-e^{-P_{3}t})$

Wherein,

C _pthe insulin concentration in blood plasma,

I ₀the quality being delivered to hypodermic insulin dose at zero time point,

D is coefficient of diffusion (insulin diffuses into the speed blood-glucose from interstitial fluid ISF),

V _pthe Plasma volumes in health,

V _iSFthe volume of the interstitial fluid ISF to its insulin delivery,

P ₂time constant,

P ₃be more than or equal to P ₂time constant, and

T is from insulin dose being sent the time entered interstitial fluid ISF.

Time constant can use quadratic formula to calculate:

$P_2,P_3=-\frac{{\mathrm{\α}}_1\±\sqrt{{\mathrm{\α}}_1^2-{4a}_0}}{2}$

Wherein,

${\mathrm{\α}}_1=\frac{D+K}{V_p}+\frac{D}{V_{\mathrm{ISF}}},$ And

${\mathrm{\α}}_0=\left(\frac{D+K}{V_P}\right)\left(\frac{D}{V_{\mathrm{ISF}}}\right)-\frac{D^2}{V_{\mathrm{ISF}}{V}_P}$

In alternative embodiments, as shown in figure 37, the lead-lag compensation device 522 that controller is rearmounted is for revising order (U _pID) to compensate insulin delivery latency and/or insulin removing speed k.The rearmounted lead-lag compensation device 522 of controller is following forms: wherein, 1/ α and 1/ γ is lead and lag constant respectively, and s is Laplace variable, U _cOMPit is the order of the compensation calculated by lead-lag compensation device 522.

PID controller produces and is used for the order (U that desirable insulin enters the delivery rate of blood plasma _pID).Order U _pIDrenewal speed based on control loop regularly calculates and sends, and this order is based on the maximum expected variation speed of blood sugar level, and the minimum insulin dose of insulin delivery system, insulin sensitivity, the minimum and maximum concentration of glucose etc. that accepts is selected.Order U _pIDas the input of the rearmounted lead-lag compensation device 522 of controller.

In specific embodiment, the compensation order (U sent from the lead-lag compensation device 522 that controller is rearmounted _cOMP) use more than one come the value of self-controller.In specific embodiment, the rearmounted lead-lag compensation device 522 of controller uses the current command (U _pID ⁿ) and previous commands (U _pID ^n-1), according to following compensation formula calculation compensation order U _cOMP:

U _COMP ⁿ＝(1-γ)U _COMP ^n-1+U _PID ⁿ+(1-α)U _PID ^n-1

Wherein,

U _pID ⁿthe current command,

U _pID ^n-1previous commands,

U _cOMP ^n-1that previous compensatory control exports,

α is with min ^-1for the leading time constant reciprocal of unit, and

γ is with min ^-1for the lag time constant reciprocal of unit.

This is the first forward difference formula.But, alternatively, other forms (such as, the first backward or bilinearity) can be used, but form of ownership all produces compensatory control exports (U _cOMP), it exports (U by history PID _pID) and history compensation output (U _cOMP) weighting historical structure.

Amendment is used for the order (U of the compensation of insulin delivery latency and/or insulin removing _pID) the weighting history that can send based on historical insulin of optional method carry out.By providing more weights of sending history recently, the weighting history that previous insulin is sent can control to deduct output from current PID subsequently, and the control be compensated exports.This is expressed as in Laplce's Main Factors:

wherein, E is the error signal (G-G that Laplce transforms _b), λ determines that the reduction that PID exports the weighting history that controls to export with history proportional is how many, and α determines that (preferred value of α can equal reciprocal and arrange time constant or subcutaneous insulin shows, P for the zero count constant of weighting how long history ₂).The compensating signal solved as the function of error obtains:

$\frac{U\left(s\right)}{E\left(s\right)}=\mathrm{PID}\frac{s+{\mathrm{\α}}_w}{s+(\mathrm{\α}+\mathrm{\λ})}=\mathrm{PID}\frac{s+{\mathrm{\α}}_w}{s+\mathrm{\γ}}$

It is identical with previously described lead-lag compensation.

In other optional embodiments, other previous commands values can be used.In other optional embodiments, compensation formula make-up time constant P ₂and P ₃.

In more Alternate embodiments, controller gain is modified to the effect comprising the rearmounted lead-lag compensator of controller, and like this, the rearmounted lead-lag compensator of controller does not need the order revising responsible insulin delivery latency.

In certain embodiments, the order of insulin delivery system response controller provides limited insulin dose to health.The minimum insulin that insulin delivery system can be sent is minimum limited insulin dose.Controller can produce the order for insulin dose to be delivered, and described dosage is not the integral multiple of minimum limited insulin dose.Therefore, response command sends too much or very few insulin by insulin delivery system.In specific Alternate embodiments, order is shortened to the immediate integral multiple of minimum limited insulin dose and adds to next order the designated volume retaining insulin by controller rearmounted lead-lag compensation device.In other optional embodiments, the immediate integral multiple of order to minimum limited insulin dose walked around by compensator.In other optional embodiments, additive method for compensate order and minimum limited insulin dose immediate integral multiple between difference.In other embodiments, do not need to compensate.

The plasma insulin feedback estimated is used to cancel lead-lag compensation device

In another optional embodiment, PID control command can be modified to imitate the effect of plasma insulin to beta cell, thus determines best insulin administration based on the plasma insulin that subcutaneous insulin infusion is estimated by feedback.The clean effect of this feedback substitutes less-than-ideal dynamics by more desirable dynamics and obtain the attainable plasma insulin curve of beta cell.This can from hereinafter finding out (using Laplace transform variable).Suppose the glucose (G-G higher than basis _b) and insulin send relation between (ID) by linear transfer function D (s)=C (s) (G (s)-G _b) describe, wherein, C (s) can be described by PID controller transfer function, but must by this function representation.If beta cell uses periphery insulin (I _p(s)) level suppression insulin secretion, the insulin delivery rate so estimated can be modified to:

D(s)＝C(s)(G(s)-G _B)-kI _p(s)

For portal vein insulin is sent, known ID (s) and plasma insulin I _ps the relation between () is by single time delay estimadon:

$I_p\left(s\right)=\frac{k_1}{s+\mathrm{\α}}\mathrm{ID}\left(s\right)$

By I _ps () value substitutes into previous formula and k is amplified, obtain:

$\mathrm{ID}\left(s\right)=\frac{C\left(s\right)\left({G\left(s\right)-G}_B\right)}{1+\frac{{\mathrm{kk}}_1}{s+\mathrm{\&alpha;}}}\&ap;C\left(s\right)\frac{s+\mathrm{\&alpha;}}{{\mathrm{kk}}_1}(G\left(S\right)-G_B);1<<\frac{{\mathrm{kk}}_1}{s+\mathrm{\&alpha;}}$

Above-mentioned formula completely eliminates undesirable time constant 1/ α.In practical operation, lower k value can be used, obtain:

$\mathrm{ID}\left(s\right)=C\left(s\right)(G\left(s\right)-G_B)-\frac{{\mathrm{kk}}_1}{s+\mathrm{\&alpha;}}\mathrm{ID}\left(s\right)=C\left(s\right)\frac{s+\mathrm{\&alpha;}}{s+\mathrm{\&gamma;}}(G\left(s\right)-G_B)$

Wherein, γ=α+kk ₁(that is, being sometimes greater than α).Therefore, time constant (γ=α+kk is faster used in the effect adding the beta cell of plasma insulin feedback ₁; γ > α) replace portal vein insulin to send time constant (α).In block diagram format:

It is equivalent to:

Send in order to this mechanism is applied to subcutaneous insulin, required is the transfer functions that sc insulin sends between plasma insulin.This transfer functions also estimated by two exponential time process (injecting response) or:

$\frac{I_p\left(s\right)}{{\mathrm{ID}}_{\mathrm{SC}}\left(s\right)}=\frac{k_2}{(s+{\mathrm{\&alpha;}}_1)(s+{\mathrm{\&alpha;}}_2)}$

Therefore,

$\begin{array}{c}\mathrm{ID}\left(s\right)=C\left(s\right)\left(G\left(s\right)\text{-}{G}_B\right)-\frac{{\mathrm{kk}}_2}{(s+{\mathrm{\&alpha;}}_1)(s+{\mathrm{\&alpha;}}_2)}\mathrm{ID}\left(s\right)\\ =C\left(s\right)\frac{1}{1+\frac{{\mathrm{kk}}_2}{(s+\mathrm{\&alpha;})(s+{\mathrm{\&alpha;}}_2)}}(G\left(s\right)-G_B)\end{array}$

In limited situation, work as kk ₂/ (s+ α ₁) (s+ α ₂) >>1 time, this is substantially equal to:

$\mathrm{ID}\left(s\right)=C\left(s\right)\frac{(s+{\mathrm{\&alpha;}}_1)(s+{\mathrm{\&alpha;}}_2)}{{\mathrm{kk}}_2}(G\left(s\right)-G_B)$

Equally, wherein, send relevant undesirable time constant with subcutaneous insulin to be eliminated.In practical operation, the velocity constant of rationality more (that is, faster time constant) can be only used to replace sending relevant undesirable time constant with subcutaneous insulin

Correct about 200 minutes of hypoglycemia drift (slowly declining)

Use PID controller previous analog beta cell to give and extend the good foresight to " first " and " second " stage insulin response in the process increasing the time period that glucose occurs.But if reducing fast appears in glucose after increasing the time period that glucose occurs, so PID controller cannot the slow decline of insulin of the lower glucose level of Accurate Prediction response.Figure 41 B illustrates the insulin response of the blood sugar level to Figure 41 A based on clinical data (being shown as data point), the correction (being shown as dotted line) that PID model (being shown as solid line) and PID drift about to hypoglycemia.

In a preferred embodiment, hypoglycemia drift controls (or bilinearity PID controller) by PD PID controller being revised as the proportional gain of using adaptability (Adaptive Proportional Gain) and corrects, and PD control is the improved form of original PID formula.As previously mentioned, Discrete PI D-algorithm is as follows:

Proportional component responds:

${P_{\mathrm{con}}}^n=K_P({\mathrm{SG}}_f^n-G_{\mathrm{sp}})$

Quadrature components responds:

${I_{\mathrm{con}}}^n={I_{\mathrm{con}}}^{n-1}+K_I({\mathrm{SG}}_f^n-G_{\mathrm{sp}});{I_{\mathrm{con}}}^0=I_b$

Derivative component responds:

${D_{\mathrm{con}}}^n=K_D\mathrm{dG}{\mathrm{dt}}_f^n$

Wherein, K _p, K _iand K _dproportional gain factor, integration gain factor and derivative gain coefficient, SG _fand dGdt _fbe filtered sensor glucose and derivative respectively, subscript n refers to discrete time.

In bilinearity PID controller, proportional gain K _pbased on integral error item.Each component is described by following formula the percentage contribution of insulin response:

${P_{\mathrm{con}}}^n={K_P}^n({\mathrm{SG}}_f^n-\mathrm{INT})$

${D_{\mathrm{con}}}^n=K_D\mathrm{dG}{\mathrm{dt}}_f^n$

${K_P}^n={K_P}^{n-1}+K_I({\mathrm{SG}}_f^n-G_{\mathrm{sp}}),$ Wherein K _p ⁰=K _p0

Wherein, proportional gain is current with speed K _i(initial value K _p0) carry out integration, and proportional component is associated with values of intercept (INT), wherein (INT<G _sp).Adaptability PD line as shown in the dotted line in Figure 39, the formula formula of improvement can be seen as just can the drift of matching hypoglycemia without the need to systematic error.

In extra embodiment, bilinearity PID controller is also multiplied by the value of such as α and so on to merge integrator seepage by amendment formula to make previous KP, as follows:

${K_P}^n=\mathrm{\&alpha;}{K_P}^{n-1}+K_I({\mathrm{SG}}_f^n-G_{\mathrm{sp}})$

Wherein, α ≈ 0.99

Correct hypoglycemia drift optional method by integrator cut out to PID contrast carry out.PID controller usually has integrator and resets rule, and it prevents excessively " winding " and this rule can be used for correcting hypoglycemia drift.Such as, integrator can be cut out as follows:

If (SG≤60mg/dl and I _con ^n-1> K _p(SP-60)), so I _con ^n-1=K _p(SP-60)

This formula resets integrator, and like this, if sensor glucose drops to lower than 60mg/dl, it is 0 that the insulin of so all sensor glucose signals stablized or decline is sent.Cut out boundary and represent absolute threshold, be similar to the counter regulation reaction of people.

But other modes more accurately can imitating beta cell comprise use piece-wise continuous function.Such as, lower array function allows progressively to cut out to regulate:

$\mathrm{\&gamma;}\left(\mathrm{SG}\right)={\mathrm{\&gamma;}}_0+(1-{\mathrm{\&gamma;}}_0)\left[\frac{T_{1-\mathrm{SG}}}{T_1-60}\right]$

If (SG≤T ₁mg/dl and $I_{\mathrm{con}}^{n-1}>\mathrm{\&gamma;}{K}_p(\mathrm{SP}-60)$ ), so $I_{\mathrm{con}}^{n-1}=\mathrm{\&gamma;}{K}_P(\mathrm{SP}-60)$

This formula introduces two extra regulating parameter (γ ₀and T ₁) and start to check that integrator exports under higher thresholds condition.Such as, if γ ₀=5 and T ₁=100mg/dl, so integrator exports and can be tailored to 4K _p60, if glucose drops to 90mg/dl, so integrator exports and can be tailored to 3Kp60, if glucose drops to 80mg/dl, etc., until glucose reaches 60, so, integrator exports and can be tailored to K _p60.Alternatively, other functions being different from the function proposed in above-mentioned formula can be used (such as, based on glucose decline rate or I _conreduce the function of number percent).

System configuration

Part below provides exemplary but infinite the illustrating of element to can be used for above-mentioned controller.Under the condition of scope not deviating from embodiments of the present invention, various change can be made to element, the layout of various different elements, the combination of element etc.

Before thinking that controller 12 provides input, sensor signal 16 is usually through Signal Regulation, and such as, pre-filtering, filtering, corrects etc.Such as prefilter, one or more than one wave filter, the element of corrector and so on separates with controller 12 or together with being physically located at, and can comprise remote measurement characteristic monitor transmitter 30, infusion apparatus 34 or utility appliance.In a preferred embodiment, as shown in Figure 8 B, prefilter, wave filter and corrector are included as a part for remote measurement characteristic monitor transmitter 30, and controller 12 is included in infusion apparatus 34.In alternative embodiments, as shown in Figure 8 C, prefilter to be included in remote measurement characteristic monitor transmitter 30 and wave filter and corrector are included in controller 12.In other optional embodiments, as in fig. 8d, prefilter can be included in remote measurement characteristic monitor transmitter 30, and wave filter and corrector can be included in utility appliance 41, and controller can be included in infusion apparatus.In order to illustrate various different embodiment in another way, Fig. 9 shows the form that the element (prefilter, wave filter, corrector and controller) in the various different equipment (remote measurement characteristic monitor transmitter, utility appliance and infusion apparatus) in Fig. 8 A to Fig. 8 D divides into groups.In other optional embodiments, utility appliance comprises some (or all these elements) in these elements.

In a preferred embodiment, sensing system produces following message, and this message comprises the information of the sensor signal based on the digital sensor value of such as digital sensor value, in advance filtering, the digital sensor value of filtering, the digital sensor value, order etc. of correction.Described message also can comprise the information of other types, such as sequence number, ID coding, check the value, value for other parameters detected, diagnostic signal, other signals etc.In certain embodiments, digital sensor value Dsig can filter in remote measurement characteristic monitor transmitter 30, the digital sensor value of filtering subsequently can be included in the message being sent to infusion apparatus 34, and wherein, the digital sensor value of filtration is corrected and in controller.In other embodiments, digital sensor value Dsig can be filtered and correct before the controller 12 be sent in infusion apparatus 34.Alternatively, digital sensor value Dsig can be filtered, correct and for controller to produce order 22, this order 22 is sent to infusion apparatus 34 from remote measurement characteristic monitor transmitter 30.

In further embodiment, such as other optional elements of rearmounted correcting filter, display, registering instrument and blood glucose meter and so on can be included in the equipment with any other element or they can be arranged separately.In general, if blood glucose meter is built up in those equipment in one, so this blood glucose meter is by the equipment that is jointly placed on corrector.In alternative embodiments, one or more than one in element is not used.

In a preferred embodiment, RF telegauge is used for the communication between equipment (such as remote measurement characteristic monitor transmitter 30 and infusion apparatus 34), and they comprise some set of pieces.In alternative embodiments, other communication medias can use between devices, such as, and electric wire, cable, IR signal, laser signal, optical fiber, ultrasonic signal, etc.

Filtering

In a preferred embodiment, the derivative of digital sensor value Dsig and/or digital sensor value is subsequently processed, filter, improve, analyze, smoothly, merging, equalization, cut out, amplify, correct etc., with the impact of minimize aberrant data point, the derivative of these digital sensors value Dsig and/or digital sensor value is supplied to controller as input subsequently.In specific embodiment, as shown in figure 16, digital sensor value Dsig, by prefilter 400, subsequently by wave filter 402, is passed to transmitter 70 subsequently.Wave filter is for detecting the impact with minimize aberrant digital sensor value Dsig.Some reasons producing abnormal digital sensor value Dsig can comprise sensor and from hypodermis, be separated produced transient signal, sensor noise, power supply noise, interim disconnection or short circuit, etc.In specific embodiment, each digital sensor value Dsig is compared with minimum and maximum value-threshold value.In other particular implementation, by digital sensor value Dsig continuous between difference compare with the change threshold speed of added value or decreasing value.

Prefilter

In specific embodiment, prefilter 400 uses fuzzy logic determination individual digit sensor values Dsig the need of adjustment.Prefilter 400 uses the subset calculating parameter in digital sensor value Dsig group and uses this parameter determination individual digit sensor values to regulate the need of relative to described digital sensor value Dsig group as a whole subsequently.Such as, the mean value of the subgroup of computable number word sensor values Dsig, is arranged on noise threshold on or below mean value subsequently.Afterwards, the individual digit sensor values Dsig in group compares with noise threshold, if the individual digit sensor values Dsig in group exceeds noise threshold, just eliminates the individual digit sensor values Dsig in group or revises.

Hereafter provide more detailed example clearly to illustrate the embodiment of prefilter, but be not limited thereto.Show the group of eight digital sensor values Dsig in fig. 17, it comprises nearest sampled value, label L, the value gathered from analog sensor signal Isig at time i and seven the preceding value K gathered from the time (i-1) to (i-7), H, G, F, E, D and C.Interior four the interim intermediate values of mean value use group (from the H that the time (i-2) gathers to (i-5), G, F and E) calculate.The mean value calculated is expressed as dotted line/dotted line average line 404.Strong noise threshold value 406 is being set up higher than average line 404100% place.In other words, the amplitude of strong noise threshold value 406 is twices of the amplitude of average line 404.Negative noise threshold 408 is being set up lower than average line 40450% place.In other words, the amplitude of negative noise threshold 408 is half of the amplitude of average line 404.The single amplitude of each in eight values (L, K, H, G, F, E, D and C) is compared with strong noise threshold value 406 and negative noise threshold 408.If value is higher than strong noise threshold value 406 or lower than negative noise threshold 408, so described value is considered to abnormal and by the amplitude replacement exceptional value of average line 404.In the example shown in Figure 17, K is higher than strong noise threshold value 406 for value, therefore uses mean value M replacement value K.And D is lower than negative noise threshold 408, therefore, with mean value N replacement value D for value.By this way, noise signal spike reduces.Therefore, in this example, value L, K, H, G, F, E, D and C are input to prefilter 400 and are worth L, and M, H, G, F, E, N and C export from prefilter 400.In alternative embodiments, other noise threshold level (or number percent) can be used.In other optional embodiments, the available value being different from mean value of value exceeding threshold value replaces, such as, preceding value, closest to the value of threshold value, the value calculated by the Trendline extrapolation of past data, by the value that interpolation between other values in threshold range calculates, etc.

In a preferred embodiment, when any one in the value in group exceeds noise threshold 406 or 408 scope, alarm flag is set.If one to three value exceeds the scope of noise threshold 406 or 408, " noise " mark is set.If exceed the scope of noise threshold 406 or 408 more than three values, so arrange " abandoning " mark, it shows that whole class value should be left in the basket and is not used.In alternative embodiments, more or less value is needed to exceed the scope of threshold value 406 or 408 to trigger " noise " mark or " abandoning " mark.

In a preferred embodiment, check the saturated of each digital sensor value Dsig and disconnect.In order to continue the example of Figure 17, each value is compared with saturation threshold 410.If value is equal to or higher than saturation threshold 410, " saturated " mark is so set.In specific embodiment, when arranging " saturated " mark, provide alarm to user, sensor 26 may need to correct or change.In further particular implementation, if each digital sensor value Dsig is equal to or higher than saturation threshold 410, so each digital sensor value Dsig can be left in the basket, and changes into the value equal to average line 404 or ignores the whole class value relevant with each digital sensor value Dsig.In a preferred embodiment, saturation threshold 410 is arranged to lower by 16% than the maximal value of the scope of issuable digital sensor value.In a preferred embodiment, maximum number sensor values represents the concentration of glucose higher than 150mg/dl.In alternative embodiments, maximum number sensor values can represent greater or lesser concentration of glucose, and this depends on the glucose concentration range of expection to be measured, sensor degree of accuracy, the sensing system resolution needed for closed-loop control, etc.The gamut of value is the difference between issuable maximum number sensor values and lowest numeric sensor values.Based on desired sensor signal scope, sensor noise, sensor gain, etc., higher or lower saturation threshold level can be used.

Similarly, in a preferred embodiment, if digital signal value Dsig is lower than disconnection threshold value 412, so arrange "off" mark, to user, this shows that sensor is not suitably connected to power supply and power supply or sensor may need to change or again correct.In further particular implementation, if digital sensor value Dsig is lower than disconnection threshold value 412, so can ignores single value, change into the value equaling average line 404, maybe can ignore the whole class value relevant with individual digit sensor values Dsig.In a preferred embodiment, disconnect threshold value 410 and be set as about 20% of gamut value.Based on the sensor signal scope of expection, sensing system noise, sensor gain etc., can use higher or lower disconnection threshold value.

In alternative embodiments, additive method is used to filter digital sensor values Dsig in advance, such as, pace of change threshold value, pace of change squared threshold, about least square fitting curve but not the noise threshold of the mean value of the subset of a class value, higher or lower noise threshold curve, etc.

Noise filter

After estimation digital sensor value Dsig, if necessary, improved by prefilter 400, make digital sensor value Dsig by wave filter 402.Wave filter 402 can be used for reducing the noise in noise, particularly frequency band.In general, the change of health blood sugar level 18 is slower relative to the speed gathering digital sensor value Dsig.Therefore, high frequency signal components is noise normally, and low-pass filter can be used to improve signal to noise ratio (S/N ratio).

In a preferred embodiment, wave filter 402 is finite impulse response (FIR) (FIR) wave filters for reducing noise.In specific embodiment, as shown in the example frequency response curve 414 in Figure 18, FIR filter is seven rank wave filters, and its passband tuned frequency is 0 to 3 circulation (c/hr) per hour and stopband tuned frequency is higher than about 6c/hr.But, usually, by frequency be 0 to as high as about 2c/hr to 5c/hr passband and effectively will reduce noise with the tuning FIR filter of the stopband 1.2 of selected band connection frequency times to 3 times when transmission sensor signal, passband tuned frequency is 0 effectively will reduce noise to as high as about 2c/hr to 10c/hr and stopband tuned frequency from 1.2 times to the 3 times FIR filter started of selected band connection frequency.In seven rank wave filters, unique weight factor is applied to each of eight digital sensor values Dsig.Digital sensor value Dsig comprises nearest sampled value and seven preceding values.Low-pass filter shows in Figure 19 A and Figure 19 B with the impact of the digital sensor value of interval collection in a minute.The unfiltered sensor signal curve 416 of digital sensor value forms distinct contrast with the curve of identical signal after seven rank FIR filter effects.Filtered signal curve 418 is delayed by and peak value is more level and smooth relative to unfiltered sensor signal curve 416.In other specific embodiments, the wave filter of more high-order or more low order can be used.In other particular implementation, based on the sensor sample speed of the expectation on body physiological basis, the computing power of remote measurement characteristic monitor transmitter 30, sensor response time etc., filter weighting coefficients can be applicable to the digital sensor value Dsig of the time interval shorter or longer than one minute collection.In alternative embodiments, based on sensor type, come from the noise of power supply or the noise of other electronic equipments, the interaction of sensor and health, body kinematics on the impact of sensor signal, etc., the wave filter with other frequency responses can be used to eliminate other noise frequencies.In other optional embodiments, wave filter is infinite impulse response (IIR) wave filter.

In alternative embodiments, additive method is used to carry out filtering to digital sensor values Dsig in advance, such as, pace of change threshold value, pace of change squared threshold, about least square fitting curve but not the noise threshold of the mean value of the subset of a class value, higher or lower noise threshold curve, etc.

Delay compensation wave filter

Except reducing noise, wave filter can be used for compensating time delay.Ideally, sensor can provide the real-time noiseless measured value of parameter, and it is such as blood sugar measured that control system is intended to control.But, in fact there is the physiology of the time delay causing measurement value sensor more delayed than blood sugar currency, chemistry, electronics and algorithm reason.

Physiology delay 422 causes the time needed for glucose moves between blood plasma 420 and interstitial fluid (ISF).Described delay is represented by the circle double-head arrow 422 in Figure 20.Usually, as discussed above, sensor 26 inserts the hypodermis 44 of health 20 and the electrode 42 near the tip of sensor 40 contacts with interstitial fluid (ISF).But ideal parameters to be measured is blood sugar concentration.Glucose is by being delivered to whole health in blood plasma 420.By diffusion process, glucose moves into the ISF of hypodermis 44 from blood plasma 420, and vice versa.Because blood sugar level 18 changes, so the glucose level in ISF changes.But the glucose level in ISF lags behind blood sugar level 18 because application reaches the time needed for the concentration of glucose balance between blood plasma 420 and ISF.Research shows that the glucose between blood plasma 420 and ISF changed retardation time between 0 to 30 minute.Some parameters that can affect the glucose retardation time between blood plasma 420 and ISF are individual metabolism, current blood glucose level, and whether blood sugar level raises or reduce, etc.

Chemical reaction postpones 424 to be introduced by the sensor response time, was represented by the circle 424 at the tip around sensor 26 in Figure 20.Sensor electrode 42 is coated with protectiveness film, and this film holding electrode 42 is soaked by ISF, reduces concentration of glucose gradually and reduces glucose fluctuation on electrode surface.Because glucose level changes, so described protectiveness film slow down the glucose exchange velocity between ISF and electrode surface.In addition, the reaction time that chemical reaction delay reacts due to glucose and glucose oxidase GOX reaction time and the secondary reaction (hydrogen-peroxide reduction is water, oxygen and free electron) producing hydrogen peroxide simply causes.

Also processing delay is there is when analog sensor signal Isig is converted into digital sensor value Dsig.In a preferred embodiment, analog sensor signal Isig carries out integration in the interval of a minute, and is converted to count numbers subsequently.In fact, A/D causes the average retardation of 30 seconds switching time.In specific embodiment, it is five minutes values that one minute value is averaged, and is sent to controller subsequently.The average retardation obtained is 2.5 minutes.In alternative embodiments, use longer or shorter integral time, cause more growing or shorter time delay.In other embodiments, analog sensor signal electric current I sig is converted to analog voltage Vsig and A/D converter every 10 seconds collection voltage Vsig continuously.Subsequently six 10 seconds values by filtering in advance and equalization to produce one minute value.Finally, five one minute values are filtered and equalization subsequently, produce five minutes values, take the photograph the average retardation causing 2.5 minutes.Other embodiments use other electronic components or other sampling rates, thus produce other of section time delay.

Wave filter also introduces delay owing to needing the digital sensor value Dsig of sufficient amount with the time run needed for wave filter.More digital sensor value Dsig is needed by the wave filter defining higher-order.Except nearest digital sensor value Dsig, FIR filter uses the preceding value with filter order equal amount.Such as, seven rank wave filters use eight digital sensor values Dsig.Generation time interval between each digital sensor value Dsig.Continue for example, if the time interval between digital sensor value Dsig is one minute, the digital sensor value Dsig the most remote so used in seven rank FIR filter be seven minutes so of a specified duration.Therefore, during average time delay for all values of wave filter 3.5 minutes.But if the unequal words of the weighting factor relevant with each value, so time delay may be longer than 3.5 minutes or shorter, and this depends on the effect of coefficient.

The preferred embodiment of the present invention comprises the various different time delays and the FIR filter higher than the high frequency noise of about 10c/hr as discussed above that compensate as discussed above up to about 30 minutes.Specific embodiment uses seven rank Weiner type FIR filter.Selected filter coefficient correction reduces high frequency noise time lag simultaneously.The example of frequency response curve 426 shows in figure 21.The example of frequency response curve 416 is 0 to 8c/hr by the band connection frequency of the sensor for sensitivity being about 20 μ A/100mg/dl and stop-band frequency produces higher than the Weiner wave filter of about 15c/hr.In dog body, carry out research by sensor and show that FIR filter can be used for compensating time delay.In research process, the time delay of wave filter for compensating about 12 minutes.The result of Figure 22 display shows the point 428 representing the actual plasma glucose level measured by blood glucose meter, and representative does not have the dotted line 430 of the measurement value sensor of delay compensation and representative to have the solid line 432 of the measurement value sensor of delay compensation.Transducer sensitivity in test is abnormal low.The research of average sensitivity sensor is used to show that the time delay of about 3 to 10 minutes is more normal in human body.The wave filter of other filter coefficients and other exponent numbers can be used for compensating time delay and/or noise.

In alternative embodiments, the wave filter of other types can be used, as long as they remove the noise of enough parts from sensor signal.In other optional embodiments, if the pace of change of blood sugar level is comparatively slow relative to time delay, so do not need time bias.Such as, within five minutes between plasma glucose and measurement value sensor, postpone not need to be corrected, play a role for closed loop glucose control system.

Derivative filter

Further embodiment can be included in before controller uses sensor signal and remove denoising from the derivative of sensor signal.Derivative is derived from digital sensor value Dsig, and it produces digital derivative sensor values (dDsig/dt).Digital derivative sensor values dDsig/dt is made to pass through FIR filter.In specific embodiment, derivative filter is at least seven rank FIR filter, and it is conditioned to remove high frequency noise.In alternative embodiments, the wave filter of more high-order or more low order can be used, and wave filter can be adjusted to the noise removing various different frequency.In other optional embodiments, derivative is available from glucose level error G _evalue, and subsequently by derivative filter 526, as shown in figure 37.In further alternative embodiment, derivative is available from analog sensor signal Isig and use hardware filter to remove denoising.

Correct

In a preferred embodiment, after filtering, digital sensor value Dsig corrects relative to one or more than one glucose reference value.Glucose reference value is inputted corrector and compares with digital sensor value Dsig.Rectifier application correcting algorithm is with converting digital sensor values Dsig, and it is counted as blood glucose value usually.In specific embodiment, bearing calibration is the U.S. Patent application the 09/511st that the name submitted on February 23rd, 2000 is called " GLUCOSE MONITOR CALIBRATION METHODS ", the type described in No. 580, this U.S. Patent application is incorporated to herein by reference.In specific embodiment, rectifier is included as a part for infusion apparatus 34, and by user, glucose reference value is inputted infusion apparatus 34.In other embodiments, glucose reference value be input to remote measurement characteristic monitor transmitter 30 and rectifier correcting digital sensor values Dsig and by correct after digital sensor value be sent to infusion apparatus 34.In further embodiment, glucose reference value is input to utility appliance, corrects in this utility appliance.In alternative embodiments, blood glucose meter is communicated with infusion apparatus 34, remote measurement characteristic monitor transmitter 30 or utility appliance, and like this, glucose reference value can be transmitted directly the equipment to being communicated with blood glucose meter.In other optional embodiments, blood glucose meter is a part for infusion apparatus 34, remote measurement characteristic monitor transmitter 30 or utility appliance, such as, the name submitted on June 17th, 1999 is called the U.S. Patent application the 09/334th of " CHARACTERISTIC MONITOR WITH ACHARACTERISTIC METER AND METHOD OF USING THE SAME ", shown in No. 996, this U.S. Patent application is incorporated to herein by reference.

In a preferred embodiment, in order to obtain blood sugar reference value, extract one or more than one blood sample from health 20, and common sales counter blood glucose meter on sale is for measuring the plasma glucose concentration of sample.Subsequently by digital sensor value Dsig with compare from the blood sugar measured of blood glucose meter, and applied mathematics correct so that digital sensor value Dsig is converted to blood glucose value.In alternative embodiments, the solution of known glucose concentrations is called the U.S. Patent application the 09/395th of " METHODAND KIT FOR SUPPLYING A FLUID TO A SUBCUTANEOUS PLACEMENTSITE " by the name such as submitted on September 14th, 1999, the method and apparatus described in No. 530 (this U.S. Patent application is incorporated to herein by reference) is introduced into the hypodermis around sensor 26, or by the solution of known glucose concentrations by using injecting method, infusion methods, jetting method, by the method that tube chamber is introduced, etc. be introduced into hypodermis around sensor 26.Collect digital sensor value Dsig sensor 26 to be soaked in the solution of known glucose concentrations simultaneously.Such as the mathematical formulae of the factor, compensation, equilibrium etc. is exported that digital sensor value Dsig is converted to known glucose concentrations.Mathematical formulae subsequently for next digital sensor value Dsig to obtain blood glucose value.In alternative embodiments, digital sensor value Dsig corrects before filtering.In other optional embodiments, digital sensor value Dsig is after filtering in advance and be corrected before filtering.In other optional embodiments, sensor is for be corrected before in body or at all without the need to correcting.

Sensor signal processing system

Before filtration and correcting, usually sensor signal is processed to convert the sensor signal of primitive form to wave filter and/or rectifier uses acceptable form.In a preferred embodiment, as shown in Figure 10, analog sensor signal Isig carries out digital quantization by A/D converter 68, and produce digital sensor value Dsig, it is sent to another equipment by transmitter 70 from remote measurement characteristic monitor transmitter 30.In a preferred embodiment, as shown in Figure 11 (a), analog sensor signal Isig is analog current value, and it is converted into the digital sensor value Dsig of digital frequency measuring value form.Universal circuit comprises integrator 72, comparer 74, counter 76, impact damper 78, timer 80 and transmitter 70.Integrator 72 produces significantly ramp voltage signal (A), and the amplitude in proportion of the instantaneous slope of ramp voltage signal and instantaneous analog sensor signal Isig.Comparer 74 converts the ramp voltage signal (A) from integrator to square-wave pulse (B).Each pulse from comparer 74 makes counter 76 increment and resets integrator 72.Timer 80 regularly triggers impact damper 78 to store currency from counter 76 and counter reset 76 subsequently.The value be stored in impact damper 78 is digital sensor value Dsig.Timer 80 also regularly can send signal to send the value from impact damper 78 to transmitter 70.In a preferred embodiment, timer period is 1 minute.But in alternative embodiments, timer period can regulate based on type of required survey frequency, sensor signal noise, transducer sensitivity, required Measurement Resolution, signal to be sent etc.In alternative embodiments, impact damper is not used.

A/D converter

Various different A/D converter designs in embodiment used in the present invention.Following Examples is exemplary, is not determinate, because can use other A/D converters.

I-F (current-frequency (counting)), single capacitor, rapid discharge

In a preferred embodiment, integrator 72 is made up of an Op-Amp92 and capacitor 82, as shown in figure 12.Integrator 72 reaches high reference voltage (VrefH) amount to analog sensor signal Isig electric current by charging to condenser voltage (A ') to capacitor 82.Condenser voltage (A ') measure under the output of a described Op-Amp92.2nd Op-Amp94 is used as comparer.When condenser voltage (A ') reaches VrefH, comparer exports (B ') paramount from low change.Of high comparator exports (B ') close the Resetting Switching 84 capacitor 82 being discharged by voltage source (V+).High capacitance exports (B ') and also triggers reference voltage switch 88 and close, almost simultaneously phase inverter 86 to make comparer export (B ') anti-phase.Phase inverter exports (C ') and triggers reference voltage switch 90 and open.Result is that the reference voltage of comparer becomes low reference voltage (VrefL) from VrefH.

When condenser voltage (A ') is discharged to VrefL, comparer exports (B ') get back to low-level, thus form pulse.Low comparer exports (B ') open Resetting Switching 84, start to make capacitor 82 again to charge.

Almost simultaneously, low capacitor exports (B ') also triggers reference voltage switch 88 and to open and phase inverter exports (C ') triggers reference voltage switch 90 and close, and this makes comparer reference voltage get back to VrefH from VrefL.

I-F, single reversible capacitance device

In alternative embodiments, use two or the polarity more than two one or more than one electric capacity of switch control rule.Embodiment shows in fig. 13.In general, one in two integrator switches 110 and 112 is only had to close and another integrator switch opens.When first integrator switch 110 cuts out, second integral device switch 112 is opened and integrator Op-Amp114 reaches high reference voltage (VrefH) by making capacitor 116 charge to condenser voltage (A ") to amount to analog sensor signal Isig current.Comparer 120 compares integrator and exports (A ") and reference voltage VrefH.And when condenser voltage (A ") reaches VrefH, comparer exports (B ") paramount from low transformation, starting impulse.

Of high comparator exports (B ") pulse and uses following method to reverse polarity of capacitor.Of high comparator exports (B ") and triggers second integral device switch 112 and close, almost simultaneously phase inverter 118 to make comparer export (B ") anti-phase.And low phase inverter exports (C "), and trigger action first integrator switch 110 is opened.Once polarity of capacitor reversing, capacitor 116 discharges with the speed proportional with analog sensor signal Isig.The reference voltage that of high comparator exports (B ") pulse also trigger comparator becomes low reference voltage (VrefL) from VrefH.When condenser voltage (A ") is discharged to VrefL, comparer exports (B ") get back to low-level.Low comparer exports (B ") open second integral device switch 112 and almost simultaneously high phase inverter export (C ") and close first integrator switch 110, this makes capacitor 116 start again to charge.Low comparer exports (B ") and trigger comparator reference voltage becomes again to VrefH from VrefL.

The advantage of this embodiment is that the sensor signal error that possible produce due to capacitor discharge time is lowered, because the amplitude of analog sensor signal Isig drives the charging and discharging speed of capacitor 116.

I-F, dual-capacitor

In further alternative embodiment, use more than one capacitor, such capacitor charges with the speed of the amplitude in proportion with analog sensor signal Isig, another capacitor discharge.The example of this embodiment shows in fig. 14.A series of three switches are used for each capacitor.First group of switch 210 is by locking (latch) voltage C " ' control, second group of switch 212 is by voltage D " ' control, it is C " ' inverse.Substantially, only have one group of switch to close at every turn.When first group of switch 210 cuts out, voltage on first capacitor 216 increases with the speed proportional with analog sensor signal Isig, until the integrator voltage of the output of Op-Amp214 (A " ') reach reference voltage (Vref).Meanwhile, one in switch makes the short circuit on the second capacitor 222, makes it discharge.Comparer 220 compares integrator and exports (A " ') and reference voltage Vref.And when integrator export (A " ') reach Vref time, comparer exports (B " ') produces pulse.Relatively export pulse and make counter 76 increment and locking output voltage C in flip-flop latch 221 " ' switch to high voltage from low-voltage.Latch voltage C " ' change cause second group of switch 212 closed and first group of switch 210 is opened.One in switch in second group of switch 212 makes the short circuit on the first capacitor 216, thus makes this capacitor discharge.Meanwhile, the voltage on the second capacitor 222 increases with the speed proportional with analog sensor signal Isig, until the integrator voltage of the output of Op-Amp214 (A " ') reach reference voltage (Vref).Again, comparer 220 compares integrator and exports (A " ') and reference voltage Vref.And when integrator export (A " ') reach Vref time, comparer exports (B " ') produces pulse.Comparer exports pulse and makes counter 76 increment and triggers locking output voltage C " ' switch to low-voltage from high voltage, this makes switch get back to its initial position, and first group of switch 210 cuts out and second group of switch 212 is opened.

In sum, when blood sugar level 18 improves, analog sensor signal Isig increases, and this causes suddenly increasing to high reference voltage VrefH fast from integrator 72 voltage out, this causes comparer 74 to produce pulse more frequently, and this makes counter 76 increase counting quickly.Therefore, the more countings of higher blood sugar level generation per minute.

Select charge storage capacity and reference voltage VrefH and VrefL of the capacitor used in integrator 72, represent to make the count resolution of the counting gathered in a minutes section under glucose level is for 200mg/dl condition the blood sugar measured error being less than 1mg/dl.In specific embodiment, VrefH is 1.1 volts and VrefL is 0.1 volt.Higher or lower reference voltage can be selected based on the measured value resolution of the capacity of the amplitude of analog sensor signal Isig, capacitor and expectation.Source voltage V+ is set to enough high, discharges so fast enough significantly not reduce the quantity of cpm under 200mg/dl blood sugar level discharge time to make one or more than one capacitor.

Pulse persistance output function

In a preferred embodiment, whenever timer 80 activating emitter 70, the digital sensor value Dsig from impact damper 78 sends by transmitter 70.But in specific embodiment, as shown in Figure 11 B, user or another individuality can use selector switch 96 to select other outputs for the treatment of to launch from transmitter 70.In a preferred embodiment, selector switch 96 is the forms of the menu shown on screen, and this menu-style is accessed individual by the button used on the surface of remote measurement characteristic monitor transmitter 30 by user or another.In other embodiments, dialing selector switch, dedicated button, touch-screen, the signal being emitted to remote measurement remote measurement characteristic monitor transmitter 30 etc. can be used.Be different from digital sensor value Dsig, can by select launch signal include but not limited to: the monopulse extended period, in advance filter before digital sensor value, in advance filter after but filter before digital sensor value, filter after digital sensor value, etc.

In specific embodiment, as shown in Figure 11 B, pulse width counter 98 counts time clock from pulse width timer 100 until pulse width counter 98 is by resetting from the rising edge of the pulse in comparer 74 or negative edge.The counting accumulated when pulse persistance counter 98 resets represents the pulse width of a part for the individual pulse of comparer 74.When reset signal trigger pulse keeping count device 98, the counting accumulated in pulse persistance counter 98 is stored in monopulse impact damper 102.When individual choice monopulse exports, the value in monopulse impact damper 102 launched by transmitter 70.The cycle of pulse persistance timer 100 must be enough shorter than the time period between the edge of the individual pulse from comparer 74, thus high analog sensor signal Isig is had be enough to the resolution that quantizes from the different pulse durations of comparer 74.

I-V (current-voltage), voltage A/D

Optional method can be used to convert analog sensor signal Isig to analog voltage signal from analog current signal.As shown in figure 15, analog sensor signal Isig is converted into analog voltage Vsig by use Op Amp302 and resistor 304.And timer 308 regularly triggers A/D converter 306 with collecting sample value from analog voltage Vsig and converts thereof into the digital signal of representative voltage amplitude subsequently.The output valve of A/D converter 306 is digital sensor value Dsig.Digital sensor value Dsig is sent to impact damper 310 and is sent to transmitter 70 subsequently.In certain embodiments, based on the expectation resolution etc. of transducer sensitivity, maximum concentration of glucose to be measured, voltage A/D converter 306, resistor 304 can be adjusted to and amplify Vsig to use the sizable part in voltage A/D converter 306 scope.

In alternative embodiments, do not need impact damper 310 and digital sensor value Dsig is directly sent to transmitter 70 from A/D converter.In other optional embodiments, digital sensor value Dsig is processed, filter, amendment, analyze, level and smooth, merging, equalization, cut out, amplify, correct etc., is sent to transmitter 70 subsequently.In a preferred embodiment, timer 308 triggered every 10 seconds and measures.In alternative embodiments, based on blood sugar level can how soon velocity variations, transducer sensitivity, control delivery system 14 new measured value needed for frequency etc., timer 308 faster or slower runs more frequently or more infrequently trigger measurement.

Finally, in other optional embodiments, as in " sensor and sensor stand " part hereafter discuss, if necessary, convert other sensor signals of the sensor from other types to digital sensor value Dsig, subsequently digital sensor value Dsig is emitted to another equipment.

Extra controller input

In general, glucose (digital sensor value Dsig) is only used as input by proportional+integral+derivative eqn (PID) insulin response controller.On the contrary, in the human body of common glucose-tolerant, healthy beta cell benefits from extra input, such as, and nerve stimulation, enteron aisle hormonal stimulation, free fatty acid (FFA) change and protein boost, etc.Therefore, in other optional embodiments, as mentioned above, one or more than one extra input can be used and PID controller is expanded.In specific Alternate embodiments, user can manual enter ancillary information, such as, carbohydrate content in the canteen, start to have meal, estimating, the beginning of sleep cycle, the section length of one's sleep of expectation, the beginning between moving period, the exercise duration estimated, the assessment of exercise intensity, etc.Subsequently, model prediction contrast function subcontrol uses supplementary to estimate the change of concentration of glucose and correspondingly to revise output order.Such as, in the individual body of NGT, before starting to have meal, nerve stimulation triggering beta cell starts excreting insulin and enters blood flow, and this is exactly before blood sugar concentration starts to raise.Therefore, in alternative embodiments, user can inform that controller is starting to have meal and controller will start excreting insulin when expecting and having meal.

In other optional embodiments, user or another individuality can manually be ignored control system or select different controller algorithms.Such as, in specific Alternate embodiments, individuality can be selected to be normalized to basal glucose level immediately, and do not use the PID controller of simulation beta cell, and use another controller, such as, there is the PID controller of different gains, for the PD controller of quick adjustment glucose, etc.Other optional embodiments allow be standardized once glucose level and estimate do not have the quadrature components just allowing individuality close PID controller of having meal.In other specific Alternate embodiments, user can select to close whole controller, thus, cuts off closed-loop system.Once closed-loop system does not control the administration of insulin, so user just can use basal rate, variable base speed, inject etc. and to programme to infusion apparatus or user can at the dosage needing make manually to input each individuality.

In other optional embodiments, measure more than one physical trait and measured value is supplied to controller as input.Physical trait measured by can being used by controller includes but not limited to: blood sugar level, blood and/or ISF pH, body temperature, amino acid in blood (comprises arginine and/or lysine, etc.) concentration, gastrointestinal hormone in blood or ISF (comprises gastrins, secretin, cholecystokinin and/or gastrin inhibitory polypeptide, etc.) concentration, other hormones in blood or ISF (comprise hyperglycemic factor, growth hormone, cortisol, progesterone and/or estrogen, etc.) concentration, blood pressure, body kinematics, respiration rate, heart rates and other parameters.

In NGT individuality, the insulin secretion of the healthy beta cell of glucose induction can be double under mistake amino acids existence condition.And, according to " Other FactorsThat Stimulate Insulin Secretion " partly (the 8th edition in Medical Physiology mono-book, write by Arthur C.Guyton, published by W.B.Saunders Company, 1991,78th chapter, the 861st page) content known, only too much occurrence of amino acid and do not have blood sugar to raise only slightly to increase insulin secretion.In specific Alternate embodiments, estimate or measure amino acid concentration, and when amino acid concentration is enough high, the insulin response of controller increases.

In NGT individuality, the gastrointestinal hormone that there is q.s in blood causes blood insulin generation foresight to increase, this dining that beta cell expect due to individuality is described and before blood sugar rising uelralante.In specific Alternate embodiments, measure or estimate the concentration of gastrointestinal hormone, and when concentration is for being enough to show to estimate that dining is so high, adjustment control order so that insulin is introduced in body, even before blood sugar level changes.In other optional embodiments, controller uses the measured value of other hormones or estimated value to regulate insulin secretion speed.

In NGT individuality, soma is ingestion of glucose in strenuous exercise's process that insulin level is significantly lower.In alternative embodiments, the physiological parameter of such as body kinematics, blood pressure, pulse, respiration rate etc. is for detecting strenuous exercise's duration section of health and providing input to controller thus, and described controller reduces (or elimination), and infusion enters the amount of the insulin of health to compensate concentration of glucose.

Sensor compensation and end-of-life detect

In certain embodiments, as shown in figure 31b, transducer sensitivity 510 can be degenerated in time.When transducer sensitivity 510 changes, sensor signal degree of accuracy reduces.If transducer sensitivity 510 significantly changes, so sensor must again correct or change.Diagnostic signal can be used to carry out evaluate sensor signal degree of accuracy whether change and/or can use diagnostic signal conditioning signal that diagnostic signal maybe can be used to represent when again to correct or more emat sensor.When transducer sensitivity 510 reduces, the glucose level 512 using sensor signal to measure underestimates actual blood sugar level 514, and the measuring error 516 between the glucose level 512 measured and actual blood sugar level 514 is increasing in time, as shown in fig. 3 ia.Transducer sensitivity 510 reduces, as shown in Figure 31 C due to the increase of sensor resistance Rs.The resistance that sensor resistance Rs is working electrode WRK and provides the health between electrode CNT, it is shown as summation or is shown as R1 and R2 in the circuit diagram of Fig. 7.Sensor resistance Rs is by measure analog sensor signal Isig and also calculate resistance (Rs=Vcnt/Isig) subsequently to electrode voltage Vcnt and indirectly obtain.

When sensor resistance Rs increases, the analog sensor signal Isig responding given concentration of glucose reduces.In a preferred embodiment, the reduction of analog sensor signal Isig is by identifying the variable quantity of the sensor resistance Rs from nearest correction and using the resistance variations in correcting algorithm 454 to regulate analog sensor signal value subsequently and be compensated.The offset calculated by correcting algorithm 454 is for increasing sensor analog signals value.When sensor resistance increases, offset increases in time.Correcting algorithm 454 comprises at least one value changed with the change of sensor resistance Rs.In certain embodiments, is correcting recently before sensor resistance Rs there occurs much changes in assessment, low-pass filter for sensor resistance Rs measurement to reduce high frequency noise.

In alternative embodiments, sensor resistance Rs can use different formulae discovery.Such as, sensor resistance Rs2 can be calculated as:

Rs ₂＝(V ₀-Vcnt/Isig)

In certain embodiments, V ₀the voltage identical with Vset.The advantage of this method is to which illustrate voltage level Vset, and this voltage levvl Vset can change with different sensors and/or change with different monitors, and/or changes with the change of analog sensor signal.This eliminates the noise relevant with the change of Vset and/or skew, and can provide more accurate sensor resistance explanation.In other particular implementation, V ₀be set to-0.535 volt, it is the common voltage of Vset.In further embodiment, V ₀calculated by paired Vcnt and Isig measured value.Use least square method or other curve-fitting methods, can obtain from the relation between Vcnt and Isig is derivative the mathematical formulae representing curve (normally straight line formula).Subsequently, V ₀by curve extrapolation is obtained to find the value of Vcnt when Isig is 0.

Figure 38 A to Figure 38 H shows and uses V ₀calculating sensor resistance and do not use V ₀comparison between calculating sensor resistance.The Rs that Figure 38 G shows ₂derivative curve more clear and more clearly show sensor fault relative to the derivative curve of Rs that Figure 38 F shows.Therefore, sensor resistance Rs ₂the sensor resistance Rs can be replaced and use or can with the sensor resistance Rs conbined usage.

In a preferred embodiment, sensor is again corrected when the change of sensor resistance Rs from correcting recently exceedes threshold value or is changed, or sensor is again corrected when the pace of change dRs/dt of sensor resistance exceedes another threshold value or changes.In certain embodiments, the pace of change dRs/dt of sensor resistance can compare with the threshold value of two shown in Figure 32.If threshold value that dRs/dt exceeds " replacing ", so sends the alarm of more emat sensor to user.If threshold value that dRs/dt exceeds " again correcting ", so sends again the alarm of correcting sensor to user.

In the example shown in Figure 33 A to Figure 33 C, analog sensor signal Isig significantly reduced about 0.3 day time, as shown in figure 33 a.When only providing analog sensor signal Isig, user can think that the reduction of analog sensor signal is the reduction due to blood sugar.But in fact, the reduction of analog sensor signal Isig causes due to the flip-flop of transducer sensitivity.Sensor resistance Rs shown in Figure 33 A increased with the reduction of analog sensor signal Isig about 0.3 day time.Spiking 522 when the derivative dRs/dt of the sensor resistance shown in Figure 33 C clearly demonstrates about 0.3 day when analog sensor signal Isig reduces.Sensor abnormality is described spiking 522 in the change of sensor resistance dRs/dt but not actual blood sugar reduces.If the threshold value on dRs/dt is set to +/-4, so user can in about alarm receiving more emat sensor for 0.3 day.As shown in figure 33 a, sensor is not until the 1.4th day changes yet.Analog sensor signal Isig from about 0.3rd day lower than true glucose levels, until about 1.4th day more emat sensor.

In certain embodiments, time dt (in this time range to sensor resistance Rs differentiate) amount be whole time from correcting recently.In other embodiments, the amount of the time dt of differentiation is fixing, such as, last hour, 90 minutes, 2 hours, etc.

In alternative embodiments, when the integration of sensor resistance Rs on schedule time window (∫ Rs d/dt) exceeds predetermined resistance integral threshold, sensor is again corrected or is changed.The advantage of this mode is that it is tending towards filtering out can from comprising the potential noise produced the signal of accidental spiking, the change of unexpected voltage levvl etc.Preferably, based in time window process with the Rs measured value that setting speed (such as, 1 minute, 5 minutes, etc.) obtains, at the integration of time window (such as 15 minutes etc.) upper calculating sensor resistance Rs.In alternative embodiments, time window can be longer or shorter and can use different sample rates, and this sample rate used based on noise, system responses, controller etc. is selected.In further embodiment, time window and sample rate can change in time, and such as, when close to the sensor life-time terminal estimated, or the sensor as indicated in formula is degenerated, etc.

As mentioned above, multiple threshold value can be used.Such as, if ∫ Rs d/dt exceeds " replacing " threshold value, the alarm of more emat sensor is so sent to user.If threshold value that ∫ Rs d/dt exceeds " again correcting ", so sends again the alarm of correcting sensor to user.In further alternative embodiment, to electrode voltage Vcnt for assessment of other features, such as, the pollution of sensor degree of accuracy, biosensor organism, sensor function, sensor voltage range of operation, sensor connect, etc.

PH controller inputs

In alternative embodiments, the measured value of both local pH of the gentle ISF around sensor of controller using-system interstitial fluid (ISF) G/W is to produce the order being used for infusion apparatus.In specific Alternate embodiments, be arranged in the Multifunction Sensor 508 of hypodermis for measuring the gentle pH of G/W.Be placed on the tip of the Multifunction Sensor 508 with three electrodes in hypodermis as shown in figure 30.Working electrode 502 is coated with platinum black and is coated with glucose oxidase (GOX).Contrast electrode 506 is coated with silver-silver chloride.And iridium oxide (Ir Ox) is coated with to electrode 504.Described by preferred sensor embodiment, analog sensor signal Isig produces at working electrode 502 place due to the reaction between glucose oxidase GOX and ISF glucose.But, in this optional embodiment, because the glucose in ISF and glucose oxidase GOX react on the working electrode (s and produce gluconic acid, therefore, reduce around the local pH in the ISF of sensor, this makes to there occurs change to the electromotive force of the iridium oxide on electrode 504 relative to contrast electrode REF.Like this, because pH reduces, so increase the voltage on electrode 504.Therefore, along with the raising of concentration of glucose, local pH reduces, and this causes increasing electrode voltage.Like this, concentration of glucose can be estimated based on to electrode voltage.Can comparing with the estimated value of the glucose level in analog sensor signal Isig electrode voltage estimated value of concentration of glucose.Two estimated values of glucose level can be used simply as check the value by weighted mean merging or an estimated value and play a role suitably to examine another detection method.Such as, if the difference between two estimated values is 10% within a period of time and this difference is increased to 50% suddenly subsequently, the prompting user sensor that so gives the alarm may need to change or again correct.

In extra Alternate embodiments, the pH level of sensor proximity can be used for detecting infection.By following the tracks of pH time dependent trend, the marked change of pH can be used for being identified in the infection that sensor proximity has occurred.Alarm is used for notifying user's more emat sensor.

PH sensor can be used in other embodiments.When insulin can do nothing to help health use glucose, health is transferred to consume fat and is provided energy.When health from use glucose convert to almost only use fat to provide energy time, the concentration of ketone acid (acetoacetate and beta-hydroxy-butanoic acid) rises from about 1mEq/ and is increased to 10mEq/ and rises so high.In specific Alternate embodiments, measure pH level with the rising of ketone acid in detection bodies.In embodiments of the present invention, alarm is provided when ISF pH level is too low to user.

The spinoff of the rising of ketone acid concentration is that sodium is extracted from the extracellular fluid of health and acid combines, and such health can drain acid.This causes hydrionic amount to increase, and this considerably increases acidosis.Serious case causes quick deep breathing, and acidosis is breathed, even dead.In other optional embodiments, ion-selective electrode (ISE) is for detecting na concn change.Special film is used for bag by ISE, and such ISE only detects the change of na concn.In specific Alternate embodiments, ISE is the 4th electrode adding to glucose sensor.In another optional embodiment, three-electrode system and silver-silver chloride reference electrode REF, Ir Ox together use electrode CNT and sodium ion selective (Na ISE) working electrode WRK.

When can appreciable impact insulin the pH measured value, end of life measured value, hormone measured value etc. of degree of accuracy sent add in controller input time, the basis of controller inputs normally glucose measurements.Glucose measurements is provided by sensing system.Further, once controller uses glucose measurements to produce order, delivery system just performs described order.Hereafter several device embodiments of sensing system and delivery system are described in detail.

Sensing system

Sensing system provides the glucose measurements used by controller.Sensing system comprises sensor, supports the sensor stand (if necessary) of sensor, remote measurement characteristic monitor transmitter, and the cable (if necessary) propagating electric power and/or sensor signal between sensor and remote measurement characteristic monitor transmitter.

Sensor and sensor stand

In a preferred embodiment, glucose sensor system 10 comprises thin film electrochemistry sensor, such as, name is called the United States Patent (USP) the 5th of " METHOD OF FABRICATING THIN FILM SENSORS ", 391, type disclosed in No. 250, the name submitted on February 10th, 2000 is called the U.S. Patent application the 09/502nd of " IMPROVED ANALYTE SENSOR AND METHOD OF MAKING THE SAME ", type disclosed in No. 204, or the thin film sensor of other types, such as, commonly assigned United States Patent (USP) the 5th, 390, No. 671, 5th, 482, No. 473 and the 5th, 586, the thin film sensor described in No. 553, above-mentioned United States Patent (USP) and patented claim are incorporated to herein by reference.Also see United States Patent (USP) the 5th, 299, No. 571.

Glucose sensor system 10 also comprises the sensor stand 28 supporting sensor 26, sensor stand such as described in following United States Patent (USP): the United States Patent (USP) the 5th that name is called " TRANSCUTANEOUSSENSOR INSERTION SET ", 586, No. 553 (also open with PCT application WO96/25088) and name are called the United States Patent (USP) the 5th of " INSERTION SET FOR A TRANSCUTANEOUSSENSOR ", 954, No. 643 (also open with PCT application WO98/56293), and name is called the United States Patent (USP) the 5th of " A SUBCUTANEOUS IMPLANTABLE SENSOR SET HAVING.THE CAPABILITY TO REMOVE OR DELIVER FLUIDS TO AN INSERTIONSITE ", 951, No. 521, these United States Patent (USP)s are incorporated to herein by reference.

In a preferred embodiment, use insertion pin 58 that sensor 26 is inserted through user's skin 46, and once sensor is placed in hypodermis 44, just takes out and insert pin 58 and abandoned.As shown in Fig. 3 C and 3D and Fig. 4, insert pin 58 and there is in the process of sensor insertion skin 46, support sensor sharp tip 59 and open trough 60.The United States Patent (USP) the 5th being called " TRANSCUTANEOUS SENSOR INSERTION SET " in name is further described to pin 58 and sensor stand 28,586, No. 553 (also open with PCT application WO 96/25088) and name are called the United States Patent (USP) the 5th of " INSERTION SET FOR A TRANSCUTANEOUS SENSOR ", 954, find in No. 643 (also open with PCT application WO 98/5629), above-mentioned United States Patent (USP) is incorporated to herein by reference.

In a preferred embodiment, sensor 26 has three electrodes 42, and as shown in Fig. 3 D and Fig. 4, three electrodes 42 are exposed to the interstitial fluid (ISF) in hypodermis 44.As shown in Figure 7, working electrode WRK, contrast electrode REF and to electrode CNT for the formation of circuit.When suitable voltage puts on working electrode WRK and contrast electrode REF, ISF provides impedance (R1 and R2) between electrode 42.Analog current signal Isig flows through health (R1 and R2, its summation is Rs) from working electrode WRK and flow to electrode CNT.Preferably, working electrode WRK is coated with platinum black and is coated with glucose oxidase, and contrast electrode REF is coated with silver-silver chloride and is coated with platinum black to electrode.The usual ground connection of voltage on working electrode WRK, and the voltage on contrast electrode REF remains on setting voltage Vset substantially.Vset is 300mV to 700mV, is preferably 535mV.

The reaction the most significantly of the voltage difference excitation between electrode is Reduction of Glucose, because first it generate gluconic acid and hydrogen peroxide (H with GOX reaction ₂o ₂).Subsequently, H ₂o ₂water (H is reduced on the surface of working electrode WRK ₂and (O O) ^-).O ^-attract the positive charge in sensor electronics, repel electronics thus and generation current.This causes the concentration of glucose in the ISF of analog current signal Isig and feeler electrode 42 proportional.Analog current signal Isig flows out from working electrode WRK, flow to electrode CNT, usually flows through wave filter and flows back into the low orbit of op-amp66.The input of Op-amp66 is setting voltage Vset.When Isig changes with the change of concentration of glucose, the Drazin inverse of op-amp66 on electrode CNT to voltage Vcnt.The usual ground connection of voltage on working electrode WRK, the voltage on contrast electrode REF is generally equal to Vset, and changes as required to the voltage Vcnt on electrode CNT.

In alternative embodiments, more than one sensor is used for measuring blood sugar.In specific embodiment, use redundant sensor.User sensor when fault is notified by remote measurement characteristic monitor transmitter electronic equipment.Indicator also can inform which sensor of user still works and/or also in the quantity of sensor of work.In other particular implementation, sensor signal is merged by equalization or other modes.If the difference between sensor signal exceedes threshold value, user is so warned again to correct or change at least one sensor.In other optional embodiments, use more than one glucose sensor, and glucose sensor not identical design.Such as, endogenous glucose sensor and exogenous glucose sensor can be used for measuring blood sugar simultaneously.

In alternative embodiments, other continuous blood sugar sensor and sensor stands can be used.In specific Alternate embodiments, sensing system is the glucose sensor that micropin analyzes thing sample devices or endogenous glucose sensor and/or use fluorescence, described micropin analyzes thing sample devices is such as called " INSERTION SET WITH MICROPIERCINGMEMBERS AND METHODS OF USING THE SAME " U.S. Patent application the 09/460th in the name of submission on Dec 13rd, 1999, the micropin describing (this U.S. Patent application is incorporated to herein by reference) in No. 121 analyzes thing sample devices, described endogenous glucose sensor is such as at United States Patent (USP) the 5th, 497, No. 772, 5th, 660, No. 163, 5th, 791, No. 344 and the 5th, 569, the endogenous glucose sensor described in No. 186, the glucose sensor of described use fluorescence is such as at United States Patent (USP) the 6th, 011, describe in No. 984, above-mentioned United States Patent (USP) is incorporated to herein by reference.In other optional embodiments, sensing system uses other detection techniques, such as, the open detection technique described in WO 99/29230 of PCT, light beam, electric conductivity, jet sampling, microdialysis, micropunch, ultrasound wave is sampled, and reverse ion permeates, etc.In other optional embodiments, only have working electrode WRK be arranged in hypodermis and contact ISF, and be positioned at external to electrode CNT with contrast electrode REF and contact skin.In certain embodiments, as shown in fig. 34 a, skin is fixed on, as a part for remote measurement characteristic monitor in surface electrode CNT and contrast electrode REF being positioned to monitor shell 518.In other particular implementation, use other equipment to be fixed on skin to electrode CNT and contrast electrode REF, other equipment shown such as, are wound around electrode with line and tie up on skin by electrode, are incorporated in by electrode on the bottom surface of the wrist-watch of contact skin, etc.In how optional embodiment, more than one working electrode WRK is placed in hypodermis, as redundancy.In extra Alternate embodiments, do not use electrode, contrast electrode REF is positioned at health external contact skin, and one or more than one working electrode WRK is arranged in ISF.Shown in Figure 34 B by example contrast electrode REF being arranged at this embodiment that monitor shell 520 is implemented.In other embodiments, from the health of individuality, ISF is obtained and the sensor external making it flow through not implant

Sensor wire

In a preferred embodiment, the type of sensor wire 32 is U.S. Patent applications the 60/121st that the name submitted on February 25th, 1999 is called " TEST PLUG AND CABLE FOR A GLUCOSE MONITOR ", the type described in No. 656, this U.S. Patent application is incorporated to herein by reference.In other embodiments, other cables can be used, such as, for propagating the low noise cable of the shielding of nA electric current, optical fiber cable, etc.In alternative embodiments, short cable can be used maybe sensor can be connected directly to equipment and without the need to using cable.

Remote measurement characteristic monitor transmitter

In a preferred embodiment, the type of remote measurement characteristic monitor transmitter 30 is U.S. Patent applications the 09/465th that the name submitted on Dec 17th, 1999 is called " TELEMETERED CHARACTERISTIC MONITORSYSTEM AND METHOD OF USING THE SAME ", the type (this U.S. Patent application is incorporated to herein by reference) described in No. 715 (being also called that the PCT application WO 00/19887 of " TELEMETERED CHARACTERISTIC MONITORSYSTEM " is open with name), and, as shown in Figure 3 A and Figure 3 B, remote measurement characteristic monitor transmitter 30 is connected with sensor stand 28.

In alternative embodiments, as shown in Figure 8 A, sensor wire 32 is connected directly to infusion apparatus shell, which eliminates the needs to remote measurement characteristic monitor transmitter 30.Infusion apparatus comprises power supply and runs the electronic component of sensor 26 storage sensor signal value.

In other optional embodiments, remote measurement characteristic monitor transmitter comprises the receiver receiving renewal or extra request for sensor data or reception and show the confirmation (manual ringing) that information has been correctly received.Specifically, if remote measurement characteristic monitor transmitter does not receive the confirmation signal from infusion apparatus, so it sends information again.In specific Alternate embodiments, infusion apparatus receives blood glucose value or other information on a periodic basis in advance.If the information of expection does not provide when asking, so infusion apparatus sends " waking up " signal and again sends information to remote measurement characteristic monitor transmitter to make it.

Insulin delivery system

Infusion apparatus

Once receive sensor signal 16 and by controller 12 processes sensor signal, the order 22 running infusion apparatus 34 will be produced.In a preferred embodiment, use external type semi-automatic medication infusion apparatus, such as, at United States Patent (USP) the 4th, 562, No. 751, 4th, 678, No. 408, 4th, 685, described in No. 903, and be called " EXTERNALINFUSION DEVICE WITH REMOTE PROGRAMMING in the name submitted on June 17th, 1999, BOLUSESTIMATOR AND/OR VIBRATION CAPABILITIES " U.S. Patent application the 09/334th, described in No. 858 (also open with PCT application WO 00/10628), above-mentioned United States Patent (USP) and U.S. Patent application are incorporated to herein by reference.In alternative embodiments, use automatic infusion apparatus, such as usual at United States Patent (USP) the 4th, 373, No. 527 and the 4th, described in 573, No. 994, this United States Patent (USP) is incorporated to herein by reference.

Insulin

In a preferred embodiment, infusion apparatus fluid reservoir 50 holds and treats that infusion enters health 20 lispro insulin.Alternatively, other forms of insulin can be used, such as, actrapid monotard, bovine insulin, pork insulin, analog or other insulin are (such as, name is called the United States Patent (USP) the 5th of " METHOD AND COMPOSITIONS FOR THE DELIVERY OFMONOMERIC PROTEINS ", 807, the name submitted in the insulin-type described in No. 315 and on January 24th, 2000 is called the U.S. Patent application the 60/177th of " MIXED BUFFER SYSTEM FORSTABILIZING POLYPEPTIDE FORMULATIONS ", the insulin-type described in No. 897, above-mentioned United States Patent (USP) and patented claim are incorporated to herein by reference), etc..In further alternative embodiment, other compositions are added in insulin, other compositions described such as, the name submitted on June 25th, 1999 is called the U.S. Patent application the 09/334th of " MULTIPLE AGENTDIABETES THERAPY ", the polypeptide described in No. 676, (name of such as submission on May 8th, 2000 is called the U.S. Patent application the 09/566th of " DEVICE ANDMETHOD FOR INFUSION OF SMALL MOLECULE INSULIN MIMETICMATERIALS " to the insulin-simulated material of Small molecular, described in No. 877), etc. (above-mentioned U.S. Patent application be incorporated to by reference herein).

Infusion tube

In a preferred embodiment, infusion tube 36 is for being transported to infusion assembly 38 by insulin 24 from infusion apparatus 34.In alternative embodiments, insulin 24 is directly transported to health 20 from infusion apparatus 34 by infusion tube.In further alternative embodiment, do not need infusion tube, such as, if infusion apparatus is connected directly to skin, so insulin directly flows to health from infusion apparatus by conduit or pin.In other optional embodiments, infusion apparatus is at body interior and infusion tube may be used for transporting insulin away from the position of infusion apparatus or can be not used in and transport insulin away from the position of infusion apparatus.

Infusion assembly

In a preferred embodiment, the type of infusion assembly 358 is types described by No. the 4th, 755,173, United States Patent (USP) that name is called " SOFTCANNULA SUBCUTANEOUS INJECTION SET ", and this United States Patent (USP) is incorporated to herein by reference.In alternative embodiments, use other infusion assemblies, such as United States Patent (USP) the 4th, 373, No. 527 and the 4th, the infusion assembly described in 573, No. 994, this United States Patent (USP) is incorporated to herein by reference.In alternative embodiments, other infusion assemblies can be used, such as, from the Rapid assembly of Disetronic, from the Silhouette of MiniMed, etc.In further alternative embodiment, do not need infusion assembly, such as, if if infusion apparatus chamber interior infusion apparatus or infusion apparatus are connected directly to skin.

With the configuration of utility appliance

In further alternative embodiment, prefilter, wave filter, rectifier and/or controller 12 are arranged in utility appliance, and this utility appliance communicates with infusion apparatus 34 with remote measurement characteristic monitor transmitter 30.The example of utility appliance comprises the U.S. Patent application the 09/487th that the name submitted on January 20th, 2000 is called " HANDHELD PERSONAL DATA ASSISTANT (PDA) WITH A MEDICALDEVICE AND METHOD OF USING THE SAME ", individual palm digital assistants described in No. 423 (being incorporated to by reference herein), computer, the module of remote measurement characteristic monitor transmitter 30 can be connected to, the module of infusion apparatus 34 can be connected to, the name that such as on June 17th, 1999 submits to is called " EXTERNAL INFUSION DEVICE WITHREMOTE PROGRAMMING, BOLUS ESTIMATOR AND/OR VIBRATIONCAPABILITIES " U.S. Patent application the 09/334th, RF programming instrument described by No. 858 (being incorporated to by reference herein) (also open with PCT application WO 00/10628), etc..In specific embodiment, utility appliance comprises rearmounted correcting filter, display, registering instrument and/or blood glucose meter.In further alternative embodiment, utility appliance comprises method user being added or revises the information to infusion apparatus 34 and/or remote measurement characteristic monitor transmitter 30 to be transmitted, such as, and button, keyboard, touch screen, etc.

In specific Alternate embodiments, utility appliance is the computer that instrument combines of programming with analyte monitoring device and RF.Analyte monitoring device receives the RF signal from remote measurement characteristic monitor transmitter 30, and they are also downloaded to computer by storage signal when needed.Control signal is sent to infusion apparatus 34 with the speed of reprogramming infusion of insulin by RF programming instrument.Analyte monitoring device and RF programming instrument are all positioned in communication station separately.Communication station comprises IR transmitter and IR receiver to communicate with analyte monitoring device and RF instrument of programming.Sensor signal value is emitted to the analyte monitoring device being arranged in communication station one by remote measurement characteristic monitor transmitter 30.Sensor signal value is communicated to computer by the IR receiver in the first communication station subsequently.Computer passes through one or more than one wave filter, corrector and controller processes sensor signal value to produce order 22.Order is sent to second communication station and is sent to RF programming instrument by the IR transmitter in communication station.Finally, RF programming instrument firing order 22 to infusion apparatus 34.Communication station, analyte monitoring device and infusion apparatus 34 can be the U.S. Patent applications the 09/409th that the name submitted on September 29th, 1999 is called " COMMUNICATION STATION FOR INTERFACING WITH AN INFUSIONPUMP; ANALYTE MONITOR; ANALYTE METER OR THE LIKE ", type described by No. 014 (also open with PCT application WO 00/18449), this U.S. Patent application is incorporated to herein by reference.Alternatively, RF programming instrument can be left in the basket and infusion apparatus can be positioned in communication station or infusion apparatus can receive order, without the need to using RF programming instrument and/or communication station.

To spend the night closed-loop system

The closed-loop insulin delivery system of type described herein can utilize various control algorithm with regulate with safety and periodically method insulin delivery to patient body.The operation of spending the night of closed-loop insulin infusion system should very carefully control in an automatic fashion, and described automated manner does not need to depend on the interaction of patient, user or paramedic.On this point, multiple safeguard procedures can be implemented with system for one day.These safeguard procedures are intended to provide available sensor glucose readings, the accuracy of estimated value sensor reading and limit insulin based on the excessive reading situation of possible sensor and send.These safeguard procedures can be warned user and make patient take suitable action.Therefore, these safeguard procedures can alleviate the potential risk of closed-loop control of spending the night.

The control algolithm that system uses can be considered to the type in safeguard procedures, because it imitates the effect of the insulin suppressing insulin secretion.This system also can implement sensor performance protection.Such as, whether closed loop starting algorithm can enter closed loop mode by calculating nearest correction factor certainty annuity.Time between the nearest correction factor of starting algorithm inspection and the previous calibration factor, and determine the relative sensors error between reading.As another example of sensor guard measure, system can during closed loop mode using forestland manager.By the sensor dextrose equivalent and real sensor dextrose equivalent comparing real-time mode prediction, described mode manager checks that sensor glucose readings is enough to use spending the night in closed loop mode.If the dextrose equivalent of model prediction and actual value are significantly different, so system triggers represents the dangerous alarm of sensor fault.This dangerous alarm can respond multiple sensors problem and produce, such as, sensor drift, sensor shift, sensor pressurized, etc.

System also can the protection of implementation goal glucose level.On this point, configurable starting algorithm is with by regulating the target glucose level in closed loop mode to provide level and smooth conversion between open loop mode and closed loop mode gradually.Target glucose after adjustment is used by closed loop control algorithm until the target glucose after regulating pools specific setting value.At this moment, described setting value can be used for the Rapid Dose Calculation in closed loop mode process in future.

System also can use at least one insulin limit to send as insulin and protect with sensor performance.In this case, the insulin limit limits the maximum amount of insulin being delivered to any time patient, thus avoid closed-loop control system due to potential sensor fault excessive insulin delivery.In practical operation, the insulin limit is to the special value of each patient and its insulin sent in fasting time section based on patient, and fasting blood-glucose and insulin sensitivity calculate.

System also can use one or more than one insulin to send protection.Such as, insulin is sent the continuous monitoring of time-out (in operation with closed ring process) patient under insulin maximum conditions, whether is received the time period that insulin continues prolongation, if so trigger dangerous alarm.Whether this protection also supervisory system does not have insulin delivery to continue the time period extended, if trigger dangerous alarm.It is that another insulin sends protection that correction is injected.If this system-computed patient under closed loop mode service condition, higher than the blood sugar threshold value of specifying, alleviates the insulin bolus dosage of hyperglycemia.Blood glucose meter reading when measuring by requiring that closed loop mode starts realizes.Correct to inject and calculate based on amount of insulin and target glucose on the insulin sensitivity of patient, plate.On plate, insulin (IOB) compensates also is that another insulin sends protection.IOB compensates and manually injects amount of insulin on estimation plate based on what give, and such system effectively can estimate IOB.On this point, manually inject and can deduct from the insulin dose calculated by PID-IFB control algolithm.

System also can implement one or more than one communication protection.Such as, the data that receive of the continuous supervisory control device of " sensor transmissions do not received " function.For amount to less than working time the packet do not received for 15 minutes for, system held closed loop mode.But within this time period, system uses closed loop control algorithm to continue to calculate insulin dose based on last effective sensor dextrose equivalent.For the packet do not received for 15 minutes to 60 minutes altogether, protection is convertible into foundation for security speed of programming in advance, is defined as the half of patient at night basal rate.If controller starts to receive the packet within the scope of foundation for security Velocity Time, system converts closed loop mode to again.For totalling over for the packet that do not receive for 60 minutes, system can convert open loop mode to, and wherein, system sends the basal rate (it can be set by paramedic) of programming in advance.

Exemplary closed loop algorithm, methodology and technology can based on the type more detailed descriptions hereinafter of the pid control algorithm provided in forward part of content disclosed by the invention.In some embodiments, closed loop control algorithm utilizes PID insulin to feed back (PID-IFB) control algolithm.More specifically, PID-IFB control algolithm and other algorithms representing extra protection, program and control together run, and described protection additionally can (and/or in other usage time intervals) application within the time period of use of spending the night.These extra protection can include but not limited to: use " the insulin limit " parameter; Based on the closed loop start-up circuit that glucose sensor corrects; Insulin (IOB) backoff algorithm on plate; Monitor the transmission do not received; And monitoring sensor glucose is relative to the change of the sensor glucose estimated.

In practical operation, should determine the best of insulin absolute rating or ideal set value.On this point, insulin absolute rating is used as the input of controller logic of each patient, and it has insulin and sends speed limit, as extra security feature to prevent because potential sensor error is by the excessive insulin delivery of controller.In some embodiments, insulin absolute rating by specifying the amount of the insulin of patient that to be delivered to during fasting, the insulin sensitivity of the fasting blood sugar of patient and patient calculates.

See Fig. 1, closed-loop system generally includes glucose sensor system 10, controller 12 and insulin delivery system 14.Although Fig. 1 describes these basic components as independent frame, system embodiment by two in illustrated frame or can merge into single physical component more than two.Such as, the research of closed-loop system detects the infusion pump (corresponding to insulin delivery system 14) configuring and can comprise traditional patient and wear, conventional continuous glucose sensor/emitter assemblies (corresponding to glucose sensor system 10), and with the mobile computing device (corresponding to controller 12) of the suitable write software application be mounted thereon.Mobile computing device can be such as: smart phone, panel computer, net book computer, digital media player, handheld video games equipment, etc.Should be understood that, desirable closed loop control function is implemented by a kind of or more than the executable program of a kind of computing machine being designed to run on a mobile computing device or application mode.Research test configurations also can comprise interpreting equipment, it is used as mobile computing device, and (it can use the standard infinite data communication technology, such as, Wi-Fi or bluetooth data communications protocol) and glucose sensor system 10 (it can use proprietary data communication protocol, its usually not and mobile computing device compatible) between data communication interface.

In other embodiments, the function of glucose sensor system 10 can be integrated in insulin delivery system 14, may as the interchangeable discardable module be connected with the shell of insulin delivery system 14.In other embodiments, the function of controller 12 can be incorporated in insulin delivery system 14, and like this, patient is without the need to carrying independently and different controller equiments.In fact, the control software design portable that controller 12 uses is arranged in insulin infusion pumps, pump monitor equipment etc., thus in those equipment, implement the function of controller 12, if this is desirable.In further embodiment, single hardware device platform can be designed to the function regulating insulin delivery system 14, glucose sensor system 10 and controller 12 suitably.These embodiments and other possible embodiments are that content disclosed by the invention considered, and the customized configuration of closed-loop system and set-up mode are not intended to the scope or the application that limit or limit Closed loop Control as herein described.

Although not display in Fig. 1, but closed-loop system can comprise conventional blood sugar instrument (such as, finger tip pierces through equipment) or can with conventional blood sugar instrument (such as, finger tip pierces through equipment) together run, described conventional blood sugar instrument provides the BG value of measurement to controller 12 and/or insulin delivery system 14, like this, recoverable glucose sensor system 10.In some embodiments, the BG value of calculating is sent to insulin delivery system 14, itself so that send BG value to controller 12, sensor calibration Summing Factor correction time.Controller 12 can process and analyze the information received, and whether enters operation with closed ring pattern with certainty annuity.On this point, controller 12 can make system enter closed loop mode before carry out the correction that checks to guarantee glucose sensor system 10 within the acceptable range.

After entering closed loop mode, insulin delivery system 14 as required with predetermined scheme (such as, interval with five minutes) send sensor glucose (SG) value to controller 12, sensor Isig value, correction factor, " insulin is sent " value, and other data.Controller 12 with under patient being maintained target glucose setting value condition, and transmits suitable control data and instruction with insulin delivery system 14 based on closed loop algorithm determination insulin dose.Insulin delivery system 14 is responsible for the insulin dose of specifying to patient delivery's controller 12.

Figure 49 is the processing module of exemplary embodiment and the block diagram of algorithm that illustrate closed-loop system controller 900, and Figure 50 is the process flow diagram of the exemplary embodiment of the control procedure 1000 illustrating the control insulin delivery system 14 that can be performed by controller 900 at least partly.Controller 12 being configured according to Figure 49 shown in Fig. 1.Figure 49 illustrates some input and output of controller 900, and wherein, parallelogram represents input, and oval expression exports, and rectangle represents the various different functional module of controller 900.In this description, " functional module " can be any process, technology, method, algorithm, the executable programmed logic of computing machine, etc.On this point, controller 900 can realize by any electronic equipment, and described electronic equipment has with the processor structure of at least one processor device and works in coordination with at least one memory component of connection with described processor structure.Described processor structure is suitably configured to perform the executable instruction of processor be stored at least one memory component, and such controller 900 can perform the various different control operation described in detail and method herein.Although Figure 49 describes multiple functional module of separating easily, but should be understood that, the allomeric function of controller 900 and configuration can be arranged alternatively and function described herein, operation and anyly can be performed by one or more than one in module as required.

The host electronic appliance of enforcement controller 900 can be used as the monitor equipment for insulin infusion devices, and wherein, described monitor equipment and insulin infusion devices are two physically separated hardware devices.In another embodiment of this system, the host electronic appliance implementing controller 900 can be used as portable wireless apparatus, wherein, and the physically separated hardware device of described portable wireless apparatus and two, insulin infusion devices room.In this case, portable wireless apparatus can be but be not limited to: mobile telephone equipment, panel computer equipment, notebook computer equipment, portable video game equipment, digital media player device, portable Medical Devices, etc.In other system embodiment, host electronic appliance and insulin infusion devices are at physics and functionally can be integrated into single hardware device.In these embodiments, insulin infusion devices can comprise the function of controller 900 as herein described.

Some embodiments of controller 900 comprise multiple synergistic function module, and it designs and is configured to determine at the insulin dose to be delivered patient maintained in operation with closed ring mode process under target glucose setting value condition that spends the night.On this point, the exemplary embodiment of controller 900 can comprise following functional module, but is not limited thereto: closed loop starts the delivery module 916 that on module 902, beginning module 904, proportional integral derivative insulin feedback (PID-IFB) control module 906, insulin limiting module 908, plate, insulin compensating module 910, insulin are sent timeout module 912, mode manager module 914 and do not received.

See Figure 50, control procedure 1000 can start wanting any time when inputting operation with closed ring pattern.Therefore, the order that control procedure 1000 can respond user's startup starts, and detects from the ruuning situation of dynamic response ordinary representation operation with closed ring (such as sleeping), etc.One or more than one systems inspection (task 1002) whether some embodiments of control procedure 1000 can be allowed to enter operation with closed ring pattern by confirmation system.This particular instance uses sensor calibration inspection, allows system to carry out closed loop mode subsequently.See Figure 49, closed loop starts module 902 and is included in task 1002.

In some embodiments, closed loop startup module 902 can consider some the sensor performance standards preventing closed loop from starting.These standards can include but not limited to: when unstable correct in (1) start-up course; (2) when there is marked change in transducer sensitivity; (3) when sensor is by potential invalid index corection, thus makes transducer sensitivity generation marked change; (4) any other situation of the mispairing between sensor in the multiple nearest correction (such as, two correct recently) of at the appointed time spacer segment and instrument can be caused.

The exemplary embodiment reception that closed loop starts module 902 is listd as input at least down: (measurement) BG value 920 of metering, at least one sensor calibration factor 922 (namely, correcting measuring value, correction data, etc.), sensor Isig value 924, and the time stamp data 926 representing correction time relevant with the sensor calibration factor 922 with BG value 920.These input data some or all of in these input data directly can be provided by insulin delivery system 14 (see Fig. 1), Changer Device, monitor equipment or any equipment in closed-loop system or are indirectly provided.This description supposes that the BG value 920 of each measurement produces the new sensor calibration factor 922 and new time stamp data 926, wherein, the sensor calibration factor 922 is associated with the correction of the glucose sensor system 10 (see Fig. 1) for monitoring patient.Specifically, the sensor calibration factor can based on gauge BG value 920 and corresponding sensor Isig value 924.

Closed loop starts module 902 and analyzes input data (currency and history value) to determine whether that permission system enters closed loop mode.Such as, closed loop startup module 902 can check the interval between two continuous correction timestamp value; Relatively recently and previous calibration factor values, etc.Closed loop starts " output " two permission patterns corresponding to system of module 902.More specifically, closed loop startup module 902 control system remains in open loop mode 928 and runs or start in closed loop mode 930.

See Figure 50, if closed loop mode is not allowed to (the "No" branch of query task 1004), so control procedure 1000 operational system, remains on open loop mode (task 1006) to make described system.On the other hand, if closed loop mode is allowed to (the "Yes" branch of query task 1004), so control procedure 1000 can start and in an appropriate manner (task 1008) start closed loop mode.See Figure 49, correction is injected 932 and can be calculated when closed loop mode starts and send (if necessary) to alleviate hyperglycemia.If the metering reading measured is greater than threshold value, so, this correction is injected 932 and is used as extra safeguard procedures to reach target blood glucose level.If control procedure 1000 is determined to need correction to inject, so produce executable suitable insulin administration instruction when closed loop mode starts by insulin delivery system.

See Figure 49, responding system can enter the determination of operation with closed ring pattern, can call and start module 904.Once system is in closed loop mode, so controller obtains the historical data that can be processed as hereafter more described in detail and use again.In some embodiments, such as, the controller acquisition data of nearest 24 hours (from insulin delivery system, from monitor, etc.).After this, controller regains packet, each sampling time section obtains sensor glucose (SG) value, sensor Isig value, the sensor calibration factor, the information relevant to the amount of the insulin sent, with send manually inject relevant information and the sensor calibration factor, but to be not limited thereto.As explained in detail below, regain Information Availability in various different safeguard procedures, and for determining final insulin dose.

Start sensor glucose (SG) value 940 of module 904 reception as input, and the function starting module 904 can respond the beginning of closed loop mode 930 and be activated (this trigger mechanism is represented by the dotted arrow 942 in Figure 49).SG value 940 is directly provided by glucose sensor system 10 or is indirectly provided by any equipment (see Fig. 1) in insulin delivery system 14, Changer Device or closed-loop system.This description hypothesis when SG value 940 becomes available, SG value 940 in a continuous manner by module 904 receive.Start module 904 and also can use target glucose setting value 944, it can be kept in inside, produced and/or provide by controller 900.For embodiment as herein described, fixing (constant) value (dotted line of Figure 49 describes target glucose setting value 944 to represent that described value is the parameter that user specifies, but not the functional module of system acceptance or data) that target glucose setting value 944 representative of consumer can specifically be specified.

In some embodiments, start module 904 and calculate final goal dextrose equivalent 946, it is used as the input of PID-IFB control module 906.Final goal dextrose equivalent 946 makes the conversion (by adjusting final goal dextrose equivalent 946 gradually) that system can be more level and smooth between open loop mode and closed loop mode.Starting module 904 can use target glucose setting value 944 to calculate final goal dextrose equivalent 946.On this point, start module 904 and final goal dextrose equivalent 946 is increased to the identical level of sensor dextrose equivalent when starting with closed loop mode, condition is that sensor glucose is higher than a certain threshold value.Along with passage of time, final goal dextrose equivalent 946 reduces gradually and is back to target glucose setting value 944 (usually in about two hours).See Figure 50, control procedure 1000 calculates final goal dextrose equivalent (task 1010) based on final goal dextrose equivalent (task 1012) at least partly and proceeds by calculating uncompensated infusion of insulin speed by PID Rate (n).For this example, task 1010 process can relate to beginning module 904 and task 1012 process can relate to PID-IFB control module 906.

As extra safeguard procedures, insulin limiting module 908 is collaborative to provide the insulin upper limit with PID-IFB control module 906, this insulin upper limit is based on the insulin intake of the patient of the fasting time section of specifying, and the fasting blood-glucose of patient and the insulin sensitivity of patient calculate.This insulin limits gives the upper limit to avoid due to the excessive insulin delivery of potential sensor error system to insulin delivery rate.

See Fig. 1 to Figure 48, PID-IFB control module 906 can be configured to the control procedure performing above-detailed.In some embodiments, PID-IFB control module 906 receives lists as input at least down: SG value 940 (it can be used for calculating the pace of change value representing SG value pace of change), current sensor Isig value 950, current sensor correction factor 952, and the amount of insulin 954 of sending.As shown in figure 49, PID-IFB control module 906 can receive the insulin limits value 959 (such as, maximum infusion of insulin speed) of the user that insulin limiting module 908 calculates.The input of PID-IFB control module 906 can by insulin delivery system 14, glucose sensor system 10, Changer Device, and any equipment in monitor equipment and/or closed-loop system provides (see Fig. 1) directly or indirectly.PID-IFB control module 906 is appropriately configured into based on current and history SG value 940, SG pace of change, sensor Isig value 950, the sensor calibration factor 952, final goal dextrose equivalent 946 and the insulin 954 sent calculate infusion of insulin speed, thus realize euglycemia.These values (and other possible values), when becoming available, can be received by PID-IFB control module 906 in a continuous manner, such as, with the interval of five minutes or according to any desirable timetable.

The insulin 954 sent represents the parameter or the value that are delivered to the amount of the insulin of patient by insulin delivery system.Therefore, the insulin 954 sent can represent the nearest bolus amount (normally some units) of sending in certain time.In some embodiments, the insulin 954 sent corresponds to the amount of insulin of sending in the nearest sampling time, and the described nearest sampling time may be, but not limited to: 1 minute, 5 minutes, 30 seconds or any sampling times of specifying.The insulin 954 sent also can represent in the past based in the fixed time section of (such as, N number of hour recently) or the amount of the insulin sent by delivery system of injecting or the amount of insulin of being sent by system in the nearest sampling period.In practical operation, the history value of the insulin 954 that PID-IFB control module 906 (with IOB compensating module 910) can be sent with Collection and preservation by " initialization " as required.After this, the insulin 954 sent can represent the amount (if by inject passage or foundation channel gives) of the insulin given by system in nearest sampling time section simply.

As mentioned above, PID-IFB control module 906 can use the insulin upper limit 959, and it is patient-specific parameter.In some embodiments, the insulin upper limit 959 can be inputted by user, paramedic etc.Alternatively, insulin limiting module 908 can be responsible for calculating or the management insulin upper limit 959 (if this expects).The insulin upper limit 959 gives the upper limit, as extra security feature, to avoid making the excessive insulin delivery of controller 900 due to potential sensor error to insulin delivery rate.Therefore, if PID-IFB control module 906 recommends the dosage higher than insulin restriction 959, massage insulin restriction 959 is used to the insulin sent to be limited to insulin limits value.In addition, the component quadrature components that the enforcement of insulin restriction 959 " can freeze " PID is to preceding value to avoid saturation integral, and described saturation integral can cause carrying out continuous integration until it reaches maximal value to glucose error.In some embodiments, the insulin upper limit 959 has the default value being set as patient base's speed five times.Therefore, if reach maximal value, so PID-IFB control algolithm can be quite positive in calculating insulin dose.Therefore, in order to minimize saturation integral, insulin restriction 959 being fed back to PID-IFB control module 906 (as described in Figure 49), calculates for next insulin dose.

PID-IFB control module 906 is run as described earlier to calculate current insulin dosage 958 as output valve (current insulin dosage 958 is also referred to as uncompensated infusion of insulin speed by PID Rate (n) in this article).In practical operation, current insulin dosage 958 is typically expressed as infusion velocity (unit/hour).When describing like this, current insulin dosage 958 can represent closed loop infusion velocity, and this speed is limited by insulin limiting module 908, and this speed can be regulated further by IOB compensating module 910 or compensate.Therefore, the output (the insulin upper limit 959) of insulin limiting module 908 represents the potential restriction insulin dose treating to be provided by PID-IFB control module 906, if do not give restriction, the output that insulin limits 959 pairs of PID-IFB control modules 906 does not produce effect, or current insulin dosage 958 is identical with the insulin upper limit 959.Refer again to Figure 50, control procedure 1000 can calculate infusion of insulin speed AdjustedRate (n) after regulating based on uncompensated infusion of insulin speed (task 1014) at least partly and compensate patient's " on plate " insulin.For this example, task 1014 process can comprise IOB compensating module 910.

IOB compensating module 910 receives lists as input at least down: current insulin dosage 958, about send manually inject 960 information.Described send manually inject 960 can by insulin delivery system 14, Changer Device, any equipment (see Fig. 1) in monitor equipment and/or closed-loop system provides directly or indirectly.This description hypothesis send manually inject 960 become available time, what receive that this sends by IOB compensating module 910 in a continuous manner manually injects 960, such as, in the interval of five minutes or according to the timetable of any expectation.IOB compensating module 910 is suitably configured to, based on the artificial bolus dose of sending, estimate insulin on plate before operation with closed ring or in process, thus compensates final infusion velocity to help to avoid the excessive insulin delivery of controller 900.Therefore, the output of IOB compensating module 910 can be the final insulin dose 962 being expressed as final infusion velocity (unit/hour).Final insulin dose 962 is also called as infusion of insulin speed AdjustedRate (n) of adjustment in this article.

See Figure 50, infusion of insulin speed AdjustedRate (n) of control procedure 1000 adjustment in use is to control insulin infusion devices, and then control procedure 1000 regulates insulin sending (task 1016) to user's body.In some embodiments, the infusion of insulin speed of adjustment communicates (such as wireless data communications) in an appropriate manner with insulin infusion devices.Control procedure 1000 can in an iterative manner or continuous print mode carry out as above with the situation of monitor user ' and insulin delivery as required, and to participate in without the need to user.That is, should be terminated (the "Yes" branch of query task 1018) if control procedure 1000 determines operation with closed ring pattern, so control procedure 1000 makes system convert back to open loop mode (task 1020).Closed loop mode can respond the order termination that user starts, and the ruuning situation that response represents open loop operation usually detects and automatically stops, etc.

Should continue (the "No" branch of query task 1018) if query task 1018 determines closed loop mode, so control procedure 1000 can check whether be in executive control program another circulation time.In other words, control procedure 1000 can check next sampling time (query task 1022).If to the time of carrying out next circulation, so control procedure 1000 can be got back to task 1010 and be repeated the calculating with next group data value.Such as, next circulation of control program can obtain and process the currency of some in following parameters or the currency of all following parameters: SG value 940, SG pace of change, sensor Isig value 924, the amount 954 of the insulin sent and send manually inject 960, but to be not limited thereto.This makes control procedure 1000 according to predetermined scheme, specifies sample rate etc., regulates final infusion of insulin speed in a continuous manner.

Whether insulin is sent timeout module 912 and is monitored patient and receiving and limit the insulin sent continuously or patient with maximum insulin and whether receiving the time that 0 unit/hour minimum admissible infusion persistence controller is specified.Therefore, insulin send timeout module 912 can receive as input the insulin 954 sent.If exceed the fixed time, so system can trigger dangerous alarm 966.Otherwise system remains in operation with closed ring pattern 968.

Get back to Figure 49, mode manager module 914 receives lists as input at least down: the insulin 954 sent, sensor Isig value 950 and one or more than one sensor calibration factor 952.The input of mode manager module 914 can by insulin delivery system 14, glucose sensor system 10, Changer Device, and any equipment (see Fig. 1) in monitor equipment and/or closed-loop system provides directly or indirectly.Mode manager module 914 is appropriately designed and is configured to the insulin 954 based on sending, and sensor Isig value 950 and the sensor calibration factor 952 in real time (or substantially real-time) estimate the concentration of glucose of user.The sensor calibration factor 952 that mode manager module 914 uses equals the sensor calibration factor 922 that closed loop starts module 902 use.That is, closed loop starts module 902 and uses the sensor calibration factor 922 at special time, and mode manager module 914 in closed loop mode operational process to carry out and continuous print mode considers the sensor calibration factor 952.When the dextrose equivalent that pattern is estimated is significantly different with sensor dextrose equivalent, system can exit closed loop mode.Therefore, mode manager module 914 regulating system remains on closed loop mode 974 and is still converted to open loop mode 976.

Leak under transmission module 916 is suitably configured to monitoring and list: sensor Isig value 950, SG value 940 and the sensor calibration factor 952, but be not limited thereto.More specifically, leak transmission module 916 and monitor whether receive with check system the packet transmitting necessary information and input value continuously.For the leakage biography packet that total is less than time threshold lower limit (such as 15 minutes), system remains in closed loop mode, as shown in the frame 980 in Figure 49.In this time durations, system uses closed loop control method to continue to calculate insulin dose based on nearest effective sensor dextrose equivalent.For amounting to higher than lower threshold and leaking the packet passed lower than the time threshold upper limit (such as, 60 minutes), leak the foundation for security speed that systematic evaluation can extremely be programmed by transmission module 916 in advance, as shown in the frame 982 in Figure 49.In some embodiments, this foundation for security speed is defined as the half of patient's overnight basal speed and this parameter can be programmed by paramedic or doctor.Start to receive packet if leak transmission module 916 with during foundation for security rate of administration, so system can switch and is back to closed loop mode.For total higher than time threshold upper limit not received data Bao Eryan, system can switch to open loop mode, as shown in the frame 984 in Figure 49.At this point, system can be sent the open loop overnight basal speed of programming in advance by control.

In sum, controller 900 responds gauge BG value 920 at least recently, and the sensor calibration factor 922 and correction time stab data 926 and determines whether to enter closed loop mode.Controller 900 uses closed loop start-up mode 902 to check whether within the acceptable range sensor calibration time between nearest two corrected values, and whether any change between two corrected values (most recent value and preceding value) is acceptable.If so, so system can be converted to closed loop mode by controller 900.Once system is in closed loop mode, so, controller 900 can periodically receive packet (such as, every five minutes), described packet comprises present SG value 940, current sensor Isig value 950, the insulin 954 sent, the sensor calibration factor 952 and send manually inject 960.In some embodiments, each packet that controller 900 receives is included in over the data of collection in 24 hours sections.

Starting module 904 uses SG value 940 and target glucose setting value 944 to calculate final goal dextrose equivalent 946.In some embodiments, target glucose setting value 944 is set to 120mg/dL, although can use other settings (if necessary (usual setting range can be such as 70-300mg/dL)).This makes by regulating the conversion that final goal dextrose equivalent 946 is more level and smooth between Open loop and closed loop pattern gradually.Final goal dextrose equivalent 946 is sent to PID-IFB control module 906, is used as an input of the calculating of the final insulin dose 962 of impact.

PID-IFB control module 906 uses final goal dextrose equivalent 946, and current and history SG value 940, SG value pace of change and the insulin 954 sent determine infusion of insulin speed (current insulin dosage 958), thus realize euglycemia.As extra safeguard procedures, from the insulin upper limit 959 of insulin limiting module 908 (based on the insulin intake of patient in fasting process, fasting blood-glucose and insulin sensitivity calculate) be input to controller 900, send speed limit to give insulin to each patient, thus avoid the excessive insulin delivery of controller 900.PID-IFB control module 906 considered the insulin upper limit 959 before transmission current insulin dosage 958 to IOB compensating module 910, this insulin upper limit 959 estimates insulin on the plate in manually injecting before operation with closed ring or in operation with closed ring process, thus calculates final insulin dose 962.Final insulin dose 962 can be sent to insulin delivery system 14 directly or indirectly from controller 900, and like this, final insulin dose 962 can be delivered to patient in operation with closed ring process.

Extra safeguard procedures can be implemented with monitoring system in operation with closed ring process, and like this, system exits closed loop mode when not meeting some standard.Such as, if the continuous data bag omitted is more than specifying number, controller 900 can make system exit closed loop mode.This hypothesis controller 900 usually receive in a continuous manner in operation with closed ring process packet (from insulin delivery system 14, monitor, conversion equipment, etc.).Therefore, if as expected, controller 900 detects that the continuous data bag higher than number of thresholds is not received, and so system is instructed to exit closed loop mode.As described earlier, this function is associated with leakage transmission module 916.

And mode manager module 914 is based on the insulin 954 sent, and sensor Isig value 950 and the sensor calibration factor 952 estimate the concentration of glucose of user in a continuous manner.If the difference between the dextrose equivalent of model prediction and sensor dextrose equivalent is greater than the threshold value of regulation, so controller 900 can make system exit closed loop mode.

As above sum up, the multiple module that controller 900 uses synergy to regulate the insulin in operation with closed ring process to send or function: closed loop starts module 902, start module 904, PID-IFB control module 906, insulin limiting module 908 and IOB compensating module 910.And controller 900 can be used in operation with closed ring process the multiple module performing various different preventing protective function.These protection modules can comprise: insulin sends timeout module 912, mode manager module 914 and leakage transmission module 916.

Closed loop starts module: first represents embodiment

Closed loop starts module 902 and checks the change of transducer sensitivity and determine whether that permission system enters closed loop mode.See Figure 49, the input that closed loop starts module 902 comprises current gauge BG value 920, and the sensor calibration factor 922 and correction time stab data 926.Closed loop starts module 902 and checks a series of condition belonging to sensor calibration factor values 922 and the time obtaining sensor calibration factor values 900.If all conditions is all satisfied, controller 900 starts operation with closed ring pattern.If do not meet this standard, so system held in open loop mode of operation and controller 900 need to carry out new sensor calibration.

Closed loop starts some embodiments (see Figure 49) of module 902 and performs one or more than one certainty annuity and whether carry out the function of closed loop mode, algorithm or method.Here is parameter and the variable that closed loop starts the exemplary embodiment use of module 902:

T=plan enters the time of closed loop mode;

Nearest correction factor (the CFR)=sensor calibration factor (CF) value recently;

TR=obtains the time of CFR;

Nearest CF value before the previous calibration factor (CFP)=CFR;

TP=obtains the time of CFP;

CFchange=for any correction factor for the number percent from previous CF to current C F change.CFchange can calculate according to following formula:

CFchange can calculate according to following formula:

CFchange=(abs (CFcurrent – CFprevious)/CFprevious) * 100 (formula 50)

TRecent=intends to start the nearest window correction time (minute) before closed loop mode

Minimum time difference (minute) between correction before tDiffmin=corrects recently and corrects recently

TDiffmax=corrects the maximum time difference (minute) between previous calibration recently

The most I of CFmin=accepts CF (mg/dL/nA)

The maximum acceptable CF of CFmax=(mg/dL/nA)

CF value before CFcurrent in CFprevious=a pair CF value

The acceptable CF of CFchengeTH=changes the threshold value (mg/dL/nA) of number percent

In some embodiments, closed loop startup module 902 performs with the form of series of processing steps.Use logic hereinafter described, closed loop starts module 902 and determines whether allow system enter closed loop mode.

Case A

If (tP is (tR-tDiffmin:tR) not in time window), so carries out following logical check.

(if CFmin≤CFR≤CFmax)

(if t-tRecent≤tR≤t)

(if tR-tDiffmax≤tP≤tR-tDiffmin)

(if CFmax≤CFP≤CFmax)

In above-mentioned formula 50, use CFR as CFcurrent

CFchange is calculated as CFprevious with CFP

(if CFchange≤CFchangeTh)

Enter closed loop

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

If any one in above-mentioned condition is not satisfied, system remains in open loop mode.Therefore, in order to enter closed loop mode, the new correction carrying out all conditions met in case A (or case B described below) can be required.

Case B

(if tP is in time window (tR-tDiffmin:tR))

Nearest CF value in CFP2=time window tR-tDiffmax:tR-tDiffmin

TP2=obtains the time of CFP2

(if CFmin≤CFR≤CFmax)

(if t-tRecent≤tR≤t)

If there is CFP2 can

In above-mentioned formula 50, use CFR as CFcurrent and use

CFP2 calculates CFchange as CFprevious

If (CFP2 between time tP2 and tR, CFR are with all

CF value in the scope of (CFmin:CFmax) and CFP2 and

CFchange≤CFchangeTh between CFR)

Enter closed loop

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

Otherwise can not closed loop be entered at that time

If any one in above-mentioned condition is not satisfied, system remains in open loop mode.Therefore, in order to enter closed loop mode, the new correction carrying out meeting all conditions in case A or case B can be required.

Start some variant embodiment of module 902 according to closed loop, system requires gauge BG and correlation-corrected when entering closed loop mode.Therefore, in these optional embodiments, closed loop starts module 902 and uses gauge BG and Isig to calculate CFR.Therefore, in some embodiments, sensor current can also be the input that closed loop starts module 902.Therefore, calculate because CFR starts module 902 self by closed loop, so the condition (that is, checking whether t-tRecent≤tR≤t) in case A and case B is always satisfied.

In certain embodiments, some in above-mentioned parameter can be fixed.On this point, fol-lowing values is used in an exemplary embodiment.Should be understood that, provide these values to be for illustrative purposes and closed loop starts the enforcement of module 902 can use different values herein, if necessary.

TRecent=120 minute

TDiffmin=120 minute

TDiffmax=480 minute

CFmin＝2.5mg/dL/nA

CFmax＝6mg/dL/nA

Closed loop starts module: second represents embodiment

According to some embodiments, the function that closed loop starts module 902 can represent as follows.Closed loop start module 902 can a series of case step form implement.On this point, first closed loop startup module 902 uses nearest gauge BG and Isig value to calculate nearest correction factor value (CFR), as shown in following formula A1:

CFR=meter BG/ (Isig-2) (formula A1)

At this, CFR is nearest correction factor value, and meterBG is gauge BG value, and Isig is sensor Isig value."-2 " in formula A1 represent the constant off-set value used by correcting algorithm when calculating correction factor and sensor glucose.

Use logic to be below used for case C or case D, closed loop starts module 902 and determines whether allow system enter closed loop mode.The condition of each case depends on the time that the nearest previous calibration factor (CFP) obtains.

Case C

The situation that the time that case C corresponds to the previous calibration before correcting recently is greater than 120 minutes.In addition, nearest correction factor (CFR) and the previous calibration factor (CFP) are in the scope of the restriction shown in following logical expression.

CFmin≤CFR≤CFmax (formula A2)

CFmin≤CFP≤CFmax (formula A3)

At this, CFR is nearest correction factor value, and CFP is previous correction value, and CFmin is the minimum value of the correction factor being set as 2.5mg/dL/nA, and CFmax is the maximal value of the correction factor being set as 6mg/dL/nA.

For case C, nearest correction time (tR) start closed loop start two hours within, as shown in following logical expression:

T – tRecent≤tR≤t (formula A4)

At this, tR is nearest correction time, and t is the time attempting entering closed loop mode, and tRecent attempts starting the nearest window correction time (being set as 120 minutes) before closed loop mode.

For case C, shown in following row logic expression formula, two little up to eight hours before the time of nearest correction factor of time (tP) of previous calibration.

TR-tDiffmax≤tP≤tR-tDiffmin (formula A5)

At this, tP is previous correction time, tR is the time obtaining CFR, tDiffmax is the nearest maximum time difference (it is set to 480 minutes (8 hours)) corrected between previous calibration, and tDiffmin corrects the minimum time difference (it is set to 120 minutes (2 hours)) between the correction before correction recently recently.

For case C, shown in following row logic expression formula, correct for variations (CFchange) is less than 35%, and wherein, CFchange calculates according to formula A6.

CFchange=(abs (CFR-CFP)/CFP) × 100 (formula A6)

CFchange≤CFchangeTh (formula A7)

At this, CFchange be for any a pair correction factor from the previous calibration factor to the correction factor of the current correction factor change number percent, CFchangeTh is acceptable CFchange threshold value (it is set to 35% for this example), CFR is nearest correction factor value, CFP be CFR before nearest correction factor value.

If the aforementioned all conditions in case C (formula A2 to A6) is all satisfied, so closed loop startup module 902 can correct the method (if necessary) of injecting by start-up simulation.But if any one condition is not satisfied, so controller 900 remains in open loop mode.Therefore, in order to enter closed loop mode, require to carry out the new correction meeting all conditions in case C or case D.

Case D

The time of the previous calibration that case D corresponds to before correction is recently less than the situation of 120 minutes.If the previous calibration before correcting recently is less than two hours, so the extra previous calibration factor (CFP2) is included in analysis.This allows closed loop startup module 902 estimated value to have the transducer sensitivity of at least two hours spans.

For case D, shown in following row logic expression formula, closed loop starts the second previous calibration factor (CFP2) that module 902 finds comparatively morning, and it is two little correction factors nearest within the scope of the time window of eight hours before the time of nearest correction factor (CFR).

TR-tDiffmax≤tP2≤tR-tDiffmin (formula A8)

At this, tP2 is the time obtaining the second previous calibration factor (CFP2), tR is the time obtaining CFR, tDiffmax be maximum time difference between tP2 and tR (for the present embodiment, it is set to 480 minutes (8 hours)), tDiffmin is the minimum time difference (for the present embodiment, it is set to 120 minutes (2 hours)) between tP2 and tR.

For case D, shown in following row logic expression formula, closed loop starts module 902 and also determines whether there is more than one correction factor (CF1 between the time and the time of nearest correction factor (CFR) of the second previous calibration factor (CFP2) ... CFn) available.

TP2≤t1...tn≤tR (formula A9)

At this, t1 ... tn observes more correction factor (CF1 ... CFn) time, tR is the time obtaining CFR, and tP2 is the time obtaining CFP2.

For case D, nearest correction time (tR) within two hours that start to start closed loop, as shown in following logical expression:

T – tRecent≤tR≤t (formula A10)

At this, tR is the time corrected recently, and t is the time attempting entering closed loop mode, and tRecent is the time window (it is set to 120 minutes for the present embodiment) attempting the nearest correction started before closed loop mode.

For case D, comprise nearest correction factor (CFR), the previous calibration factor (CFP), the second previous calibration factor (CFP2) and CF1 ... all correction factors of CFn are in the limited field shown in following logical formula.

CFmin≤CFR≤CFmax (formula A11)

CFmin≤CFP≤CFmax (formula A12)

CFmin≤CFP2≤CFmax (formula A13)

CFmin≤CF ₁... CF _n≤ CFmax (formula A14)

At this, CFR is nearest correction factor, and CFP is the previous calibration factor, and CFP2 is the second previous calibration factor, CF ₁cF _nit is the correction factor obtained between tP2 and tR, CFmin is the minimum value (it is set to 2.5mg/dL/nA for the present embodiment) of correction factor, and CFmax is the maximal value (it is set to 6mg/dL/nA for the present embodiment) of correction factor.

For case D, the correct for variations (CFchange) between CFR and CFP2 is less than 35%, and shown in following row logic expression formula, wherein, CFchange calculates according to formula A15.

CFchange=(abs (CFR-CFP2)/CFP2) × 100 (formula A15)

CFchange≤CFchangeTh (formula A16)

At this, CFchange is from the previous calibration factor to the correction factor of current correction factor change number percent for any a pair correction factor, CFchangeTh is acceptable CFchange threshold value (it is set to 35% for the present embodiment), CFR is nearest correction factor value, and CFP2 is the nearest correction factor value in the time range of formula A7 description.

If aforementioned all conditions is all satisfied for case D (formula A7 to A14), so closed loop startup module 902 can correct the method for injecting (if necessary) by start-up simulation.But if any one condition is not satisfied, so controller 900 remains in open loop mode.Therefore, in order to enter closed loop mode, the new correction carrying out meeting all conditions in case C or case D can be required.

The correction using IOB to compensate is injected

As mentioned above, correct inject 932 can the execution when starting to carry out closed loop mode.Correct the insulin that the object of injecting is to be provided for the hyperglycaemia alleviated when closed loop mode starts.This realizes by obtaining blood sugar gauge reading before startup closed loop first immediately.If BG gauge reading value is higher than a certain corrected threshold (CTH, it is 180mg/dL for the present embodiment), controller 900 can based on the insulin sensitivity (ISF of patient, mg/dL/ unit), insulin and desirable target glucose level (TG on plate, mg/dL) insulin delivery dosage, thus make the blood sugar of patient reach target glucose level.

According to some embodiments, correction is injected the gauge BG value (in units of mg/dL) obtained when (CB) starts based on closed loop mode and is sent, as follows:

$\mathrm{CB}=\left\{\begin{array}{cc}\frac{\mathrm{BG}-\mathrm{TG}}{\mathrm{ISF}}-\mathrm{IOB}\left(n\right),& \mathrm{if}\mathrm{}(\mathrm{BG}>\mathrm{CTH})\\ 0,& \mathrm{if}\mathrm{}(\mathrm{BG}\&le;\mathrm{CTH})\end{array}\right.$ (formula A17)

At this, CB corrects to inject, BG is blood sugar gauge value (mg/dL), TG is target glucose (mg/dL), ISF (see formula A18) is the insulin sensitivity factor (mg/dL/ unit) that patient regulates, CTH is blood sugar corrected threshold (mg/dL), exceed this threshold value can send correction and inject, and IOB (n) is from biologically active insulin (unit) on the plate of manually injecting, wherein, n is foregoing current sampling point.

ISF=ISFfactor × ISF ₀(formula A18)

At this, ISF is the insulin sensitivity factor (mg/dL/ unit) that patient regulates, ISF ₀be the insulin sensitivity factor (mg/dL/ unit) set up of patient and ISFfactor is ISF regulatory factor (without unit).The default value of ISFfactor is set as 1, and it makes ISF=ISF ₀.But for this specific embodiment, ISFfactor has been designated as the adjustable parameter in 0.5 to 2 scope, thus provide greater flexibility for optimizing the insulin sensitivity factor of patient.

$\mathrm{CB}=\left\{\begin{array}{cc}\mathrm{CB},& \mathrm{if}\mathrm{}(\mathrm{CB}\&GreaterEqual;0)\\ 0,& \mathrm{if}\mathrm{}(\mathrm{CB}<0)\end{array}\right.$ (formula A19)

At this, CB corrects to inject metering, is expressed as " unit ".Should be understood that, use formula A19, inject because controller 900 can only send forward.

Start module

Start module 904 processes sensor glucose (SG) value 940 and target glucose setting value 944 (it is set to 120mg/dL in some embodiments), thus calculate final goal dextrose equivalent 946, itself so be used as the input of PID-IFB control module 906.Therefore, final goal dextrose equivalent 946 is sent to PID-IFB control module 906, thus calculates final insulin dose 962.Refer again to Figure 49, the beginning of response closed loop mode, start module 904 and be " activated " or start.

When operation with closed ring pattern starts, beginning module 904 calculates the difference between present SG value 940 and target glucose setting value 944, as shown in following formula 51.

$\mathrm{DeltaGlu}\left(n\right)=\left\{\begin{array}{cc}\mathrm{SG}\left(n\right)-\mathrm{Setpoint},& m=1\\ 0,& m>1\end{array}\right.$ (formula 51)

In formula 51, SG is nearest sensor dextrose equivalent, n is current sampling point, Setpoint is user-defined target glucose setting value, m is sampling time in operation with closed ring process (m=1 shows that closed loop mode starts, and m increases along with each sample of receiving in closed loop mode process).DeltaGlu (n) is forced to 0 under the following situations described by m>1 and formula 52.

If DeltaGlu is less than a certain threshold value (being called MinDeltaGlu) of setting in controller 900, so starts module 904 and DeltaGlu is forced to 0.Or if DeltaGlu is greater than the threshold value of setting in controller 900, so DeltaGlu (n) keeps its value, described by formula 52:

$\mathrm{DeltaGlu}\left(n\right)=\left\{\begin{array}{cc}0,& \mathrm{DeltaGlu}\&le;\mathrm{MinDeltaGlu}\\ \mathrm{DeltaGlu}\left(n\right),& \mathrm{DeltaGlu}>\mathrm{MinDeltaGlu}\end{array}\right.$ (formula 52)

At this, DeltaGlu is the difference between the target glucose setting value 944 of present SG value 940 and definition, it is calculated by above-mentioned formula 51, further, MinDeltaGlu is the minimal difference (setting in controller 900) allowed between present SG value 940 and target glucose setting value 944.

Dynamic setting value DynSP (n) calculates according to discrete order transfer function model.As shown in formula 53:

DynSP(n)＝cd ₁·DynSP(n-1)+cd ₂·DynSP(n-2)+cn ₀·DeltaGlu(n)+cn ₁·

DeltaGlu (n-1) (formula 53)

At this, DynSP is dynamic setting value, and n is current sampling point, and n-1 is nearest sampled point, and n-2 is time nearest sampled point.Parameter cd ₁, cd ₂, cn ₀and cn ₁it is the coefficient of setting value model.These parameters are based on two time constant (τ of setting value model _sp1and τ _sp2) calculate, as follows:

cd ₁＝eaxx1+eaxx2

cd ₂＝-eaxx1×eaxx2

cn ₀＝1

${\mathrm{cn}}_1=\frac{(\mathrm{axx}1\&times;\mathrm{eaxxl}-\mathrm{axx}2\&times;\mathrm{eaxx}2)}{\mathrm{daxx}21}$

Wherein:

axx1＝1/τ _sp1

axx2＝1/τ _sp2

eaxx1＝e ^-axx1·Ts

eaxx2＝e ^-axx2·Ts

daxx21＝axx2-axx1

In above-mentioned formula, Ts represent by minute in units of sampling interval duration, τ _sp1and τ _sp2it is the time constant of fixed point.And axxl is time constant τ _sp1inverse, axx2 is time constant τ _sp2inverse, eaxx1 is τ _sp1the exponential damping factor, eaxx2 is τ _sp2the exponential damping factor, daxx21 is τ _sp1reciprocals sums τ _sp2inverse between difference.

Final goal dextrose equivalent 946 adds target glucose setting value 944 by dynamic setting value (calculating in formula 53) and calculates, as shown in formula 54:

FinalTarget (n)=Setpoint+DynSP (n) (formula 54)

In certain embodiments, some beginning in the above-mentioned parameter of module 904 fixing.On this point, fol-lowing values can be used in the embodiment of exemplary.Should be understood that, these values only provide with illustrational object at this, and the embodiment starting module 904 can use different values, if necessary.

Setpoint＝120mg/dL

MinDeltaGlu=30mg/dL (routine); 0mg/dL (lower limit); 600mg/dL (upper limit)

τ _sp1=25 minutes (routine); 0.1 minute (lower limit); 250 minutes (upper limit)

τ _sp2=30 minutes (routine); 0.1 minute (lower limit); 250 minutes (upper limit)

PID-IFB control module

PID-IFB control module 906 is based on current sensor dextrose equivalent and historical sensor dextrose equivalent 940, sensor Isig value 950, sensor dextrose equivalent pace of change, the sensor calibration factor 952, final goal dextrose equivalent 946, target glucose setting value 944, insulin restriction (such as the insulin upper limit) and the insulin 954 sent calculate infusion of insulin speed (final insulin dose 962), thus realize euglycemia.In some embodiments, PID-IFB control module 906 receives its input for every five minutes, and the calculating of insulin feedback component considers that controller 900 receives the frequency of input data thus.

PID-IFB control module 906 calculates current insulin dosage 958 according to formula 55:

U (n)=P (n)+I (n)+D (n)-γ ₁i _sC-γ ₂i _p-γ ₃i _eFF(formula 55)

It should be noted that PIDRate (n) ≡ u (n).In formula 55, infusion of insulin speed u (n) represents the current insulin dosage 958 shown in Figure 49.In formula 55, P (n), I (n) and D (n) are ratio, the anomalous integral derivative component of PID controller respectively.Insulin feedback component corresponds to its remainder.Variable γ ₁, γ ₂and γ ₃represent adjustment factor.According to some embodiments, γ ₁=0.64935, γ ₂=0.34128 and γ ₃=0.0093667, although other values can be used.Parameter I _sc(n), I _p(n) and I _eFFn (), corresponding to the state of Insulin Pharmacokinetics model, it corresponds respectively to subcutaneous, blood plasma and effective site chamber.Therefore, the insulin concentration estimated in the amount of the insulin sent and different chamber reduces pro rata.

The exemplary embodiment of the control algolithm that PID-IFB control module 906 uses was implemented in discrete (sampling) time.In the symbol used, n is current time step, wherein, and=nT _s, and at use T _sduring sampling time section, t be continuous time (by minute in units of).

Proportional component P (n) calculates as follows:

P (n)=Kp [SG (n)-Final Target (n)] (formula 56)

At this, Kp be whole controller gain (with unit/hour/mg/dL represents), SG (n) is current sensor glucose, and n represents current sampling point, and FinalTarget (n) is the final goal glucose setpoint calculated from formula 54.Should be understood that, Kp is patient-specific value, and therefore, the actual value of Kp is different with the difference of patient.Although scope can change based on patient, at great majority in typical case, the value of Kp can in the scope of 0.008 to 0.200.

Quadrature components I (n) can calculate as follows:

$I\left(n\right)=I(n-1)+\frac{K_p\&CenterDot;T_s}{T_I}[\mathrm{SG}\left(n\right)-\mathrm{FinalT}\arg \mathrm{et}\left(n\right)]$ (formula 57)

At this, I (n-1) is the quadrature components from prior sample point, and n represents current sampling point, and n-1 represents prior sample point, K _pbe whole controller gains, Ts is sampling time section, T ₁be integration time constant, SG (n) is current sensor glucose, and FinalTarget (n) is the final goal glucose setpoint calculated by formula 54.

Derivative component D (n) can calculate as follows:

D (n)=K _p× T _d× dSGdt (n) (formula 58)

At this, K _pwhole controller gains, T _dbe derivative time constant, dSGdt (n) is the derivative of sensor dextrose equivalent (cross in advance and filter noise), and n represents current sampling point.

The parameter needing the controller 900 setting (adjustment) is K _p, τ _iand τ _d(see formula 56,57 and 58).For first three pid parameter, they can carry out according to formerly research hereinafter described calculating (these produce good controller performance in first method).Conventional (that is, for the situation not having insulin to feed back) controller gain K _p0based on TDI I every day of patient _tDD(unit/sky) calculates, represented by formula 59:

$K_{P0}=\frac{60}{\left(90\right)\left(1500\right)}{I}_{\mathrm{TDD}}$ (formula 59)

At this, K _p0failed controller gain, I _tDDit is TDI every day of the patient in units of unit/sky.

The object of insulin feedback is to allow controller 900 to send more insulin (such as early stage, when occurring to have meal interference), but prevent Superfusion insulin to inject with preventing of using in existing insulin pump the mode like estimator compute classes of injecting superposed.Therefore, when using insulin feedback, controller gain K _pcan be conditioned to make the insulin delivery system under steady state (SS) (that is, basal delivery situation) to send the insulin of the amount identical with regular situation.This is by conventional controller gain K _p0(not having insulin feedback to calculate) is multiplied by (1+ γ ₁+ γ ₂+ γ ₃) realize, as follows:

K _p=K _pfactorK _p0(1+ γ ₁+ γ ₂+ γ ₃) (formula 60)

At this, K _poverall controller gain, K _pfactor is K _pgain factor, K _p0failed controller gain, γ ₁(0.64935) be the adjustment factor of subcutaneous insulin concentration, γ ₂(0.34128) be the adjustment factor of plasma insulin concentrations, and, γ ₃(0.0093667) be the adjustment factor of effective insulin concentration.

Quadrature components I (n) is also equipped with antisaturation and saturation degree restriction to solve saturation integral problem.This divides shear force to realize (upper limit of quadrature components, IClip), as shown in following formula by calculated product:

$\mathrm{IClip}\left(n\right)=\left\{\begin{array}{cc}I\max \left(n\right),& (\mathrm{SG}\left(n\right)>\mathrm{UnwindHigh})\\ \mathrm{Iramp}\left(n\right),& (\mathrm{SG}\left(n\right)>\mathrm{UnwindLow})\mathrm{and}(\mathrm{SG}\left(n\right)<\mathrm{UnwindHigh})\\ \mathrm{Ilow}\left(n\right),& (\mathrm{SG}\left(n\right)<\mathrm{UnwindLow})\end{array}\right.$ (formula 61)

Imax (n)=Imaxfactor × Basal × (1+ γ ₁+ γ ₂+ γ ₃) (formula 61a)

At this, Imaxfactor is the gain factor of Imax, and Basal is the basal rate at night of patient.According to these expression formulas, when sensor dextrose equivalent is greater than upper limit threshold (UnwindHigh), IClip equals Imax (it is constant value).In some embodiments, the value of Imax can reach about 15.In typical case, Imax can have the default value of 5.0.When sensor dextrose equivalent is between upper limit threshold and lower threshold (UnwindLow), IClip equals Iramp (n), and its carrying out as shown in formula 62 calculates:

$\begin{array}{c}\mathrm{Iramp}\left(n\right)=\mathrm{Kp}\&CenterDot;[\mathrm{SG}\left(n\right)-\mathrm{UnwindLow}]\&CenterDot;\left(\frac{\mathrm{SG}\left(n\right)-\mathrm{UnwindLow}}{\mathrm{UnwindHigh}-\mathrm{UnwindLow}}\right)\&CenterDot;(I\max -\mathrm{Kp}\&CenterDot;\\ [\mathrm{Setpoint}\left(n\right)-\mathrm{UnwindLow}])\end{array}$

(formula 62)

At this, Kp is overall controller gain, SG (n) is current sensor glucose, UnwindLow is sensor glucose lower threshold, UnwindHigh is sensor glucose upper limit threshold, Imax is constant value, and Setpoint (n) is user-defined target glucose setting value.

Finally, if sensor glucose is lower than UnwindLow threshold value, so IClip supposes the value of Ilow (n), and it calculates by formula 61:

Ilow (n)=K _p[Setpoint (n)-Unwindlow] (formula 63)

At this, Kp is overall controller gain, and Setpoint is user-defined target glucose, and UnwindLow is sensor glucose lower threshold.

Figure 51 is the figure that IClip according to an example (unit/hour) changes with sensor glucose level (mg/dL).Figure 51 describes the relation between Imax, Ilow, UnwindLow and UnwindHigh of this particular instance.

The quadrature components I (n) that formula 57 calculates must be less than or equal to the IClip value shown in formula 64:

$I\left(n\right)=\left\{\begin{array}{cc}\mathrm{IClip}\left(n\right),& I\left(n\right)>\mathrm{IClip}\left(n\right)\\ I\left(n\right),& I\left(n\right)\&le;\mathrm{IClip}\left(n\right)\end{array}\right.$ (formula 64)

Insulin feedback component corresponds to its remainder.As mentioned above, for this particular instance, γ ₁=0.64935, γ ₂=0.34128, and γ ₃=0.0093667 (adjustment factor), simultaneously parameter I _sC(n), I _p(n) and I _eFFn (), corresponding to the state of Insulin Pharmacokinetics model, they correspond respectively to subcutaneous, blood plasma and effective site chamber.Therefore, the amount of the insulin sent is to reduce pro rata with the insulin concentration in the different chamber estimated.

The model describing Insulin Pharmacokinetics (insulin PK) is provided by following formula:

I _sC(n)=α ₁₁× I _sC(n-1)+β ₁× I _d(n) (formula 65)

Wherein:

${\mathrm{\&alpha;}}_{11}=\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s})$ (formula 65a)

${\mathrm{\&beta;}}_1=\left(\frac{60}{1}\right)\&CenterDot;[1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s})]$ (formula 65b)

At this, Ts is sampling time (it is five minutes in this example), τ _sbe estimation subcutaneous insulin level by minute in units of time constant (it is set to 50 minutes in this example).

I _p(n)=α ₂₁× I _sC(n-1)+α ₂₂× I _p(n-1)+β ₂× I _d(n) (formula 65c)

Wherein:

${\mathrm{\&alpha;}}_{21}={\mathrm{\&tau;}}_s\&CenterDot;[\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s})-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})]/({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)$ (formula 65d)

${\mathrm{\&alpha;}}_{22}=\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})$ (formula 65e)

${\mathrm{\&beta;}}_2=\left(\frac{60}{1}\right)\&CenterDot;({\mathrm{\&tau;}}_s\&CenterDot;[1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s})]-{\mathrm{\&tau;}}_p\&CenterDot;[1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})\left]\right)/({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)$ (formula 65f)

At this, Ts is the sampling time, and it is 5 minutes in the present embodiment, τ _sbe estimation subcutaneous insulin level by minute in units of time constant, it is set to 50, τ in the present embodiment _pbe estimation plasma insulin level by minute in units of time constant, it is set to 70 in the present embodiment.

I _EFF(n)＝α ₃₁×I _SC(n-1)+α ₃₂×I _P(n-1)+α ₃₃×I _EFF(n-1)+β ₃×I _D(n)

(formula 66)

For formula 66:

$\begin{array}{c}{\mathrm{\&alpha;}}_{31}={\mathrm{\&tau;}}_s\&CenterDot;[{\mathrm{\&tau;}}_s\&CenterDot;({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)\&CenterDot;\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s})-{\mathrm{\&tau;}}_p\&CenterDot;({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)\&CenterDot;\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})+{\mathrm{\&tau;}}_e\&CenterDot;({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)\&CenterDot;\\ \exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_e})]/[({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)\&CenterDot;({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_e)\&CenterDot;({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)]\end{array}$

(formula 66a)

${\mathrm{\&alpha;}}_{32}={\mathrm{\&tau;}}_p\&CenterDot;[\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_e})]/({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)$ (formula 66b)

${\mathrm{\&alpha;}}_{33}=\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_{\mathrm{e\&theta;}}})$ (formula 66c)

$\begin{array}{c}{\mathrm{\&beta;}}_3=\left(\frac{60}{1}\right)\&CenterDot;\left(\right[{\mathrm{\&tau;}}_s^2\&CenterDot;({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)\&CenterDot;(1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_s}))]-{\mathrm{\&tau;}}_p^2\&CenterDot;({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_e)\&CenterDot;[1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_p})]+{\mathrm{\&tau;}}_e^2\&CenterDot;\\ ({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)\&CenterDot;[1-\exp (-\frac{\mathrm{Ts}}{{\mathrm{\&tau;}}_e})])/({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_p)\&CenterDot;({\mathrm{\&tau;}}_s-{\mathrm{\&tau;}}_e)\&CenterDot;({\mathrm{\&tau;}}_p-{\mathrm{\&tau;}}_e)\end{array}$

(formula 66d)

At this, Ts is the sampling time, and it is 5 minutes in the present embodiment, τ _sbe by minute in units of the time constant of subcutaneous insulin level of estimation, it is set to 50, τ in the present embodiment _pbe by minute in units of the time constant of plasma insulin level of estimation, it is set to 70 in the present embodiment, and τ e be by minute in units of the time constant of interstitial fluid insulin level of estimation, it is set to 55 in the present embodiment.

ID (n) calculates and the infusion of insulin given.

Symbol (n-1) represents previous time step.

I _sCit is subcutaneous insulin model assessment/prediction.

I _pit is plasma insulin model assessment/prediction.

I _eFFit is the insulin model estimation/prediction of effective site.

For this particular instance, insulin PK model parameter α ₁₁, α ₂₁, α ₂₂, α ₃₁, α ₃₂, α ₃₃, β ₁, β ₂and β ₃be set to 0.9802,0.014043,0.98582,0.000127,0.017889,0.98198,1.1881,0.0084741 and 0.00005 respectively.These values are calculated by the PK-PD data of insulin, the part empirically studied and investigate.Should be understood that, particular value provided herein only reflects one group of possible desired value, and any one or more than one being adjusted in these values is suitable for particular implementation.

Insulin limits

The final infusion velocity calculated by formula 55 is restricted to described speed and is no more than the maximum insulin upper limit (Umax), as shown in following formula 67:

$u\left(n\right)=\left\{\begin{array}{cc}U\max ,& u\left(n\right)>U\max \\ u\left(n\right),& u\left(n\right)\&le;U\max \end{array}\right.$ (formula 67)

The calculating of Umax as shown in following formula 68:

$U\max =I_{\mathrm{basal},0}+\frac{{\mathrm{BG}}_{\mathrm{LBL}}-{\mathrm{FBG}}_0}{\mathrm{KI}}$ (formula 68)

At this, Umax is no more than BG _lBLthe maximum infusion of insulin speed (see following formula 68a) of (see following formula 67a), I _{basal, 0}make the fasting blood-glucose of patient reach FBG ₀the user-defined basal rate of value.BG _lBL(mg/dL) be buffering BG lower limit when reaching Umax, FBG ₀be the blood sugar of the metering glucose readings estimation used at the end of night, and KI is the insulin yield value calculated by following formula 69.

BG _lBL=Setpoint-ILB (formula 68a)

At this, BG _lBL(mg/dL) be buffering BG lower limit when reaching Umax, Setpoint is target glucose setting value 944 (Figure 49), it is defined by the user, ILB is the restriction of insulin buffering, and it is the amount (mg/dL) that system allows the extra buffering of the more insulin requirements of process.Such as, the ILB of 50 allows system to send extra insulin so that Setpoint is reduced 50mg/dL.

$\mathrm{KI}=-\mathrm{IS}*3\left(\frac{\mathrm{mg}}{\mathrm{dL}}\mathrm{per}\frac{U}{h}\right)$ (formula 69)

At this, KI is insulin gain, and IS is by the insulin sensitivity of 1800 rule estimations, and " 1800 rule " is IS=1800/TDI (total insulin every day).In some embodiments, Umax value calculates based on each patient, and the typical range of Umax be 0.5 to 3.0 unit/hour.

And if insulin restriction 959 is effective, so the quadrature components of PID-IFB algorithm is frozen in its preceding value.This characteristic can be used for helping prevent saturation integral.This information is passed and is back to PID-IFB control module 906 and uses in next PID computation process.

$I\left(n\right)=\left\{\begin{array}{cc}I(n-1),& u\left(n\right)=U\max \\ I\left(n\right),& u\left(n\right)<U\max \end{array}\right.$ (formula 70)

As mentioned above, previously research can change for configuration and/or the object running PID-IFB control module 906.First research (research 1; The people such as Panteleon, 2006) investigated and changed controller gain under study for action to the impact of eight kinds of diabetes dogs.For an experiment, conventional gain calculates based on TDI every day, and in bipartite experiment, conventional gain increases and reduces by 50 percent.The insulin of six hours sections is sent and is trended towards increasing with the increase of gain after the meal, but this does not reach statistical conspicuousness.This is because send with the lower insulin of reduction feedback of glucose level.In fact, gain is higher, more trend towards providing more insulin (realizing much better glucose to regulate) earlier and lowering later stage infusion, but, gain is lower, more trend towards infusion of insulin to remain on the time period of on foundation level lasting one period longer, because glucose level maintains apparently higher than target.It should be noted that controller gain affect pid algorithm institute important, comprise integration item and derivative term.

PID controller is applied to ten human subject by second research (studying the people such as 2, Steil, 2006).In this study, the gain of conventional controller ratio of gains conventional controller provided herein is high by 42%, integration time constant is identical (therefore, still higher integration response is had), but derivative time constant slightly lower (from rising or the definition of reduction blood sugar aspect, but not responding at daytime or night).In suggestion embodiment provided herein, night, the time constant of derivative was only lower slightly than the time constant used in research 2.

PID controller is applied to 17 human subject by the 3rd research (studying the people such as 3, Weinzimer, 2008).For the subgroup of eight patients, do not give to inject before the meal, and in all the other nine patients, gave standard meal at about 15 minutes before the meal and inject about 50% inject.Controller regulates identical with provided herein.In both cases, performance is acceptable, injects before the meal and contributes to reducing peak plasma glucose fluctuation after the meal.When with when using the house of standard pump therapy treat to compare, two kinds of closed loop algorithms are all very excellent, and its reduction is higher than 180mg/dl and the glucose oscillation lower than 70mg/dl.

One in the observations of research 3 is that after having meal, dining Related Insulin infusion maintains higher than persistent levels before the meal at least four hours.Insulin feedback is introduced in algorithm by this result, for compensating the insulin sent, allows to have an effect more energetically in case of need when starting to have meal simultaneously.

In certain embodiments, can be fixed for some in the above-mentioned parameter of PID-IFB control module 906.On this point, fol-lowing values can be used in exemplary embodiment.Should be understood that, these values only provide with illustrational object in this article, and if if required, the enforcement of PID-IFB control module 906 can use different values.

γ ₁＝0.64935

γ ₂＝0.34128

γ ₃＝0.0093667

α ₁₁＝0.90483741803596

α ₂₁＝0.065563404170158

α ₂₂＝0.931062779704023

α ₃₁＝0.00297495042963571

α ₃₂＝0.083822962634882

α ₃₃＝0.083822962634882

β ₁＝5.70975491784243

β ₂＝0.202428967549153

β ₃＝0.202428967549153

The value of above-mentioned pid parameter (KP, τ I and τ D) is not expected and changes, but if it is desirable that the value changing above-mentioned pid parameter can improve the response of glucose oscillation, so just regulates the value of above-mentioned pid parameter.The conventional value of these pid parameters and exemplary admissible scope, list in table 2.

The adjustable parameter of table 2:PID-IFB control module

Controller gain K _p0lower limit can make controller gain K _p0reduce by 95% relative to routine, this makes the overall active response of controller 900 less and sends less insulin to the gentle trend of identical G/W.The upper limit makes gain be increased to conventional 20 times.Although this can cause giving more insulin, this is usually relevant to actual glucose level and derivative thereof, thus reduces blood sugar fast, even if under the condition that blood sugar level is higher, causes insulin to send reduction due to derivative component.

Integration time constant τ _idetermine the deviation how accelerated accumulation of blood sugar and desirable blood glucose target.High value produces lower response, thus maintains the deviation of glucose and target.Finally, derivative time constant τ _dcan decreasing value zero safely always, because derivative time constant τ _dlower, the insulin sent is fewer.Any derivative time constant more much higher than the upper limit of 500 minutes can make controller too sensitive and with the speed of sensor glucose change, very little change cannot occur, even if this is not necessarily dangerous, but says that this is undesirable from overall performance.

IOB compensating module

Figure 52 is the block diagram illustrating a kind of suitable embodiment describing IOB compensating module 910 in further detail.As outlined, (namely IOB compensating module 910 regulates current insulin dosage 958 as required, the infusion of insulin speed provided by PID-IFB control module 906) to produce final insulin dose 962 (that is, the infusion velocity of the final adjustment of insulin infusion devices use).In order to calculate final insulin dose currency 962 and/or in order to future value 962, the IOB compensating module 910 calculating final insulin dose also can receive as input basal rate data 990 and with send manually inject 960 relevant information.Basal rate data 990 can represent the current basal speed of the insulin being delivered to user, and manually injecting of sending 960 can represent each amount of injecting delivering medicine to user, and time stamp data corresponds to each date of delivery/time of injecting.On this point, manually injecting of sending 960 can comprise any amount of information of injecting of sending in the past, and manually injecting of sending 960 can respond injecting of at every turn sending forward as required and upgrade.And basal rate can carry out dynamic adjustments, if need or expect (automatically being regulated by system, user, paramedic etc.).

That sends manually injects 960 Collection and preservations that can be associated with IOB compensating module 910, as injecting history 992.On this point, any amount of history bolus amount that history 992 can be included in administration in section preset time is injected.IOB compensating module 910 also can use multiple constant, parameter, coefficient, configuration setting, yield value, etc. (in brief, constant 994 shown in Figure 52 is intended to comprise these constants, parameter, coefficient, configuration setting, yield value, etc. and can be used to calculate by IOB compensating module 910 any other amount of final insulin dose 962).The IOB history 996 that Figure 52 also describes, it represents the IOB value (that is, the history IOB value of sampling time calculating in the past) previously calculated.As explained in detail below, IOB history 996 and inject the determination that history 992 can affect final insulin dose 962.Should be understood that, inject history 992, constant 994 and IOB history 996 can be stored and remain in the mnemonic element of one or more than one host computer system.For the simple and clear object with being easy to describe, Figure 52 shows these data item of IOB compensating module 910 " inside ".

IOB compensating module 910 provides extra safeguard procedures, and these safeguard procedures by activating insulin on the plate of manually injecting estimation patient body sent before closed loop mode, thus compensate final insulin dose 962 and help to avoid excessive insulin delivery.When system starts to enter closed loop mode, IOB compensating module 910 considers manually injecting 960 and deducting and manually inject the sending of having given of time period (such as, past eight hours) limited.After this, during closed loop mode, what the adjustment of IOB compensating module was sent in each sampling period (such as, every five minutes) manually injects.

Figure 53 is the process flow diagram of the exemplary embodiment illustrating IOB compensation process 1100, and this IOB compensation process can be performed by IOB compensating module 910.Process 1100 represents the Zhou Xunhuan performed at current sampling point or time n.Therefore, process 1100 receives, and obtains or access multiple input (task 1102) that can have an impact to the output of IOB compensating module.Such as, process 1100 can use the currency of uncompensated infusion of insulin speed by PID Rate (n), and as PID-IFB control module 906 produces, this input is also referred to as current insulin dosage 958 in this article.As required, process 1100 also can use current basal speed (being provided by basal rate data 990), injects some in some and/or the IOB history 996 in history 992.

When process 1100, if the IOB of activity is greater than a certain threshold value, IOB compensating module 910 manually injects insulin on computing board and compensating controller output speed (infusion of insulin speed) by each circulation.Therefore, process 1100 calculates, and produces or obtain the current I OB value (task 1104) of the estimated value of representative of consumer activity in vivo insulin.In some embodiments, active IOB is according to discrete three Room Insulin Pharmacokinetics (PK) model assessments, as follows:

IOB(n)＝ci ₁·IOB(n-1)+ci ₂·IOB(n-2)+ci ₃·IOB(n-3)+cb ₀·Ubolus(n)+

cb ₁·Ubolus(n-1)+cb ₂·Ubolus(n-2)

(formula 71)

At this, IOB is biologically active insulin on plate, and Ubolus is the amount (unit/sample) of manually injecting of sending, and n is current sampling point, and n-1 is last sampled point, and n-2 is penultimate sampled point, and n-3 is third from the bottom sampled point.Therefore, process 1100 is at least partly injected delivering data based on the history of user and is obtained current I OB value, IOB (n), and (see sending in Figure 52 manually inject 960 with inject history 992).Parameter ci ₁, ci ₂, ci ₃, cb ₀, cb ₁and cb ₂it is the coefficient of absorption of insulin model.These parameters are based on three time constant (τ of Insulin Pharmacokinetics model _sc, τ _pand τ _eff) calculate, as follows:

Ci ₁=eaxx3+eaxx4+eaxx5 (formula 71a)

Ci2=-(eaxx3 × eaxx4+ (eaxx3+eaxx4) × eaxx5) (formula 71b)

Ci ₃=eaxx3 × eaxx4 × eaxx5 (formula 71c)

cb ₀＝1

cb ₁＝dprod×(-(daxx22×eaxx3+daxx22×eaxx4)×axx3×axx4+(daxx31×

eaxx3+daxx31×eaxx5)×axx3×axx5-(daxx32×eaxx4+daxx32×eaxx5)×

axx4×axx5)

(formula 71d)

cb ₂＝dprod×（daxx22×eaxx3×eaxx4×axx3×axx4+daxx32×eaxx4×

eaxx5×axx4×axx5-daxx32×eaxx3×eaxx5×axx3×axx5)

(formula 71e)

Wherein:

Axx3=1/ τ _sC(formula 71f)

Axx4=1/ τ _p(formula 71g)

Axx5=1/ τ _eff(formula 71h)

Eaxx3=e ^-axx3TsC(formula 71i)

Eaxx4=e ^-axx4TsC(formula 71j)

Eaxx5=e ^-axx5TsC(formula 71k)

Daxx22=axx4-axx3 (formula 71l)

Daxx31=axx5-axx3 (formula 71m)

Daxx32=axx5-axx4 (formula 71n)

Dprod=-1/ (daxx22 × daxx31 × daxx32) (formula 71o)

In above-mentioned formula, TsC represents the sampling interval (minute) of amendment, and it can be calculated as TsC=Ts*6/CurveSpeed, and wherein, Ts is sampling interval and CurveSpeed is insulin rate of decay (hour) on plate.τ _sc, τ _pand τ _effsubcutaneous, the blood plasma of insulin PK model and the representative time constant of effective chamber.

The IOB (unit) that formula 71 calculates represents that described manually injecting must be considered for calculating final insulin delivery rate from manually injecting biologically active insulin residual in the body of (described manually injecting can give before closed loop mode starts or during operation with closed ring).First this by calculating IOB speed (task 1106) and deducting IOB speed to realize subsequently from the infusion velocity that PID-IFB calculates, as follows.Therefore, in some cases, process 1100 determines the infusion of insulin speed (task 1108) of adjustment at least partly based on the IOB speed calculated and uncompensated infusion of insulin speed by PID Rate (n).

$\mathrm{IOBRate}\left(n\right)=\left\{\begin{array}{cc}\mathrm{GainIOB}\&times;\mathrm{IOB}\left(n\right),& \mathrm{IOB}\left(n\right)>\mathrm{MinIOB}\\ 0& \mathrm{IOB}\left(n\right)\&le;\mathrm{MinIOB}\end{array}\right.$ (formula 72)

AdjustedRate (n)=max (0, PIDRate (n)-IOBRate (n)) (formula 73)

It should be noted that process 1100 is at least partly based on current I OB value, IOB (n), calculates IOB speed IOBRate (n).The amount of the biologically active insulin by manually injecting accumulation in time per unit body is represented with the IOB speed that unit/hour (U/h) expresses.Therefore, this extra insulin be present in body is deducted by from controller delivery rate (PIDRate).Which illustrate given by user allly manually inject and minimize the excessive possibility of sending of controller.At this GainIOB, to be the IOB rate of decay (h-1), MinIOB be needs the minimum IOB (wherein, MinIOB is with unit representation) compensating PIDRate.Therefore, when current I OB value is greater than minimum IOB value, IOB speed is calculated as and equals current I OB value and be multiplied by the IOB rate of decay, and when current I OB value is less than or equal to minimum IOB value, IOB speed is calculated as and equals 0.On this point, MinIOB is the minimum threshold of IOB, and lower than this threshold value, the effect of IOB to glucose is considered to insignificant, does not therefore need to compensate.

As formula 73 reflect, process 1100 selects the infusion of insulin speed regulated to be difference (from task 1106) between maximum or uncompensated infusion of insulin speed and the IOB speed of calculating.It should be noted that the difference between PIDRate and IOBRate can be negative, because these values are calculated by different sources.PIDRate is the infusion velocity that calculates of controller and IOBRate is from manually injecting the biologically active insulin accumulated in the body that obtains.Therefore, formula 73 guarantees that AdjustedRate is not less than 0.

Following process 1100 can calculate, selects or determine final infusion of insulin speed (task 1110).In some embodiments, final infusion of insulin speed (the final insulin dose 962 in Figure 49) as follows calculating like that:

$\mathrm{FinalRate}\left(n\right)=\left\{\begin{array}{cc}\max (\mathrm{Basal},\mathrm{AdjustedRate}\left(n\right)),& \mathrm{PIDRate}>\mathrm{Basal}\\ \mathrm{PIDRate}\left(n\right),& \mathrm{PIDRate}\&le;\mathrm{Basal}\end{array}\right.$ (formula 74)

As shown in this formula, process 1100 selects the infusion of insulin speed (AdjustedRate (n)) regulated, and uncompensated infusion of insulin speed (PIDRate (n)) or current basal speed (Basal) are as the final infusion of insulin speed (FinalRate (n)) of insulin infusion devices.At this, PIDRate is the infusion of insulin speed that PID-IFB control module 906 calculates, and Basal is current pump foundation speed of programming in advance.Therefore, when current basal speed is more than or equal to uncompensated infusion of insulin speed, task 1110 selects the final infusion of insulin speed equaling uncompensated infusion of insulin speed.On the contrary, when current basal speed is less than uncompensated infusion of insulin speed, task 1110 selects the infusion of insulin speed (being as the criterion using the higher person) of current basal speed or adjustment as final infusion of insulin speed.

When task 1110, PIDRate is used as FinalRate (when PIDRate is less than or equal to Basal) to allow controller " apply the brakes " (in other words, suppress insulin delivery rate), thus prevent any potential hyperglycaemia.On the other hand, when PIDRate is greater than Basal, FinalRate can be maximum Basal or maximum AdjustedRate, which ensure that insulin regulates only for from the insulin of injecting, but is not used in basal insulin.When PIDRate is greater than Basal, lower limit (that is, the value of Basal) is applied to FinalRate; This lower limit is for preventing insulin on overcompensation plate in these cases.

Process 1100 is by being sent to final infusion of insulin speed FinalRate (n) or being provided to insulin infusion devices and proceeding (task 1112).For the embodiment that process 1100 is performed by insulin infusion devices self the machine, process 1100 can be sent control module to the processing logic of infusion apparatus or liquid stream simply and be provided final infusion of insulin speed.And then insulin infusion devices is by regulating sending of insulin to respond according to final infusion of insulin speed.

This description hypothesis process 1100 repeated in each sampling time.Therefore, for next sampling time, n value can increase next circulation onset index (task 1114) that 1 (or increasing any desirable amount) thinks process 1100.After this, process 1100 can return task 1102 to obtain up-to-date input data values and to repeat above-mentioned various different task.Therefore, when system is just run in closed loop mode, process 1100 regulates insulin delivery to deliver to user's body in mode that is controlled and that carry out continuously by regulating final infusion of insulin speed to promote continuously.

In some embodiments, it is desirable that some in the parameter regulating IOB compensating module 910 to use, can improving SNR if done like this.Conventional value and the exemplary admissible scope of these parameters are listed in table 3.

Parameter	Conventional value	Lower limit	The upper limit
				CurveSpeed	6	1	8
GainIOB	1.25	0	5
				MinIOB	1	0	500

The customized parameter of table 3.IOB compensating module

Insulin sends timeout module

Insulin is sent timeout module 912 and is designed appropriately and is configured to continuous monitoring (during closed loop mode) patient and whether under insulin restriction (Umax) condition or under not having the condition of insulin (Umin, it may be defined as considerably less to not having insulin to send (unit/hour)), receive the time period that insulin continues one period longer.If these insulin one of sending in condition is detected, so system can give the alarm and under closed loop mode continuous service.As previously mentioned, insulin send timeout module 912 can process as input the insulin 960 sent.

Therefore, insulin is sent timeout module 912 and is introduced extra safeguard procedures, and it checks and is used under insulin restriction (Umax time-out) condition or insulin delivery continues the longer time period under not having the condition of insulin (Umin time-out) a series of following step.This is by calculating the insulin total amount that during closed loop mode, system is sent in the preassigned Moving Window being identified as insulin time window.

About Umin Timeout conditions, once arrive the insulin time window of Umin (it can be defined as insulin delivery under 0 unit/hour condition) from closed loop mode, the amount of the insulin that system is sent under can monitoring the insulin time window of specifying user and it being compared with the amount of having sent when running under patient base's speed of same time span, shown in following row logic expression formula:

$\begin{array}{c}\mathrm{PumpDeliveryRate}=\\ \left\{\begin{array}{c}\mathrm{FinalRate},\mathrm{if}{U}_{\mathrm{Tot}}^{\mathrm{WinMin}}>(\mathrm{MinDeliveryTol}/100)\&times;U_{\mathrm{Basal}}^{\mathrm{WinMin}}\\ \mathrm{Alert},\mathrm{if}{U}_{\mathrm{Tot}}^{\mathrm{WinMin}}\&le;(\mathrm{Min\; Deliv\; eryTol}/100)\&times;U_{\mathrm{Basal}}^{\mathrm{WinMin}}\end{array}\right.\end{array}$ (formula 75)

At this, Pump Delivery Rate (unit/hour) be infusion velocity, it equals FinalRate in formula 74 (that is, the infusion velocity that controller calculates during closed loop mode) or equals the overnight basal speed of programming in advance that uses in open loop operation process. amount be the insulin total amount (unit) that in user's Umin insulin time window of specifying, closed loop control algorithm is sent, and amount be run and the insulin total amount of sending with the overnight basal speed of programming in advance in identical Umin insulin time window.Parameter MinDeliveryTol is the tolerance that user specifies, and it must be sent to maintain system in closed loop mode number percent.

According to this particular instance, as long as the insulin total amount that system is sent in insulin time window (being set as 120 minutes in this example) process be greater than on basis 5 percent (for this example, it is fault minimum metric) can the total amount of transmissible insulin when running under condition, so closed-loop control proceeds.And, once the insulin total amount that system is sent in insulin time window process is less than 5 percent of basis, with regard to trigger fault safety alarm.

About Umax Timeout conditions, once arrive the insulin time window of Umax from closed loop mode, it also runs with Umax speed with under same time span and can the amount of transmissible insulin compare, shown in following row logic expression formula by the amount of the insulin sent under the system insulin time window condition that just monitoring is specified user:

$\begin{array}{c}\mathrm{PumpDeliveryRate}=\\ \left\{\begin{array}{c}\mathrm{FinalRate},\mathrm{if}{U}_{\mathrm{Tot}}^{\mathrm{WinMin}}>(\mathrm{MinDeliveryTol}/100)\&times;U_{\mathrm{Basal}}^{\mathrm{WinMin}}\\ \mathrm{Alert},\mathrm{if}{U}_{\mathrm{T\; ot}}^{\mathrm{WinMin}}\&le;(\mathrm{MinDeliveryTol}/100)\&times;U_{\mathrm{Basal}}^{\mathrm{WinMin}}\end{array}\right.\end{array}$

(formula 76)

At this, Pump Delivery Rate is infusion velocity, and it equals the overnight basal speed of programming in advance used in FinalRate or operational process in an open loop mode. amount be the insulin total amount (unit) that in the Umax insulin time window of specifying user, closed loop control algorithm is sent, and the amount Moving Window of specifying user when being and running under the Umax speed calculated in the insulin total amount that may send.Parameter MaxDeliveryTol is the tolerance that user specifies, and it must maintain in order to system under remaining on closed loop mode number percent.

According to this particular instance, as long as the insulin total amount that in insulin time window (being set as 600 minutes in this example) process, system is sent be less than under 95% (for this example, it is fault maximum metric) condition of Umax run and can transmissible insulin total amount.And running under 95% condition that the insulin total amount of sending once system in insulin time window (600 minutes) process is greater than Umax can transmissible insulin total amount, so with regard to trigger fault safety alarm.

In some embodiments, it is desirable that adjustment insulin sends some in the parameter of timeout module 912 use, if if doing so, can improving SNR.Conventional value and the exemplary admissible scope of these parameters are displayed in Table 4.

Parameter	Conventional value	Lower limit	The upper limit
				The Umin of insulin time window	120 minutes	30 minutes	600 minutes
The Umax of insulin time window	600 minutes	30 minutes	600 minutes
				MinDeliveryTol	5％	0％	100％
MaxDeliveryTol	95％	0％	100％

Table 4. insulin sends the adjustable parameter of timeout module

Mode manager module

Mode manager module 914 is similar to insulin and sends timeout module 912, because mode manager module is monitored and supervised the system in operation with closed ring process.In practical operation, closed-loop system only knows the signal (input) provided by measuring equipment.If measured value departs from actual value, so control system can react to departing from.When using the continuous glucose sensor of diabetic, sensor provides measured value to closed-loop control system, and based on these measured values, insulin is delivered to patient.Therefore, sensor performance and integrality should be subject to close monitoring.Fortunately, produce between the insulin of blood sugar response and meal ingestion and there is association.This association can be translated into can based on the mathematical model of the insulin prediction sensor glucose responding sent.The sensitivity of sensor glucose for the insulin sent is patient-specific (for each patient, sensitivity can be learnt usually in three days values process of six days).

Model manager module 914 uses the mathematical model of the plasma glucose time dependent response personalization that patient can be made exclusive.Described model describes the sensor glucose time dependence reaction as the function of insulin and meal ingestion.Exemplary mathematical model described herein has a lot of benefit and advantage: it is linear, based on physiological and only comprise and have with measurable data (sensor glucose and the insulin sent) parameter contacted directly.These characteristics are extremely important, because linear model is easy to analysis and prediction.And, be conducive to based on physiological model source (such as, the insulin sensitivity understanding prediction, the picked-up of meals, etc.) and use measurable data to reduce (such as, the metabolism of estimation nonobservable variable, cytoactive, etc.) demand.

Figure 54 is the figure of definition for some time-events of model management.Mark " present " represents that nearest sampling time or sampling period 1120, k equal the time period that the current sampling time deducts the length (LPH) corresponding to estimation range in the sampling time.Figure 54 also represents the time period of the length (LTH) corresponding to training area in sampling, and it becomes the model training phase.Insulin history is defined as the length of the data estimated needed for plasma insulin.In order to make model manager pattern 914 estimate fault, consider that in the past insulin history adds the record of the insulin sent in LTH and LPH sampling time section, and eight ten at least percent of Isig (electric signal) measured value from k-LTH and k.

As described in detail below, model manager module 914 considers " Moving Window " that comprise historical time section 1122, and described historical time section is defined as from (nearest sampling time section 1120) gets back to LTH at present.The Moving Window that model manager module 914 is considered also can comprise the insulin history prior to LTH, described by Figure 54.The data obtained in each time window process are processed and analyze near current time or time at present, are preferably processed before next sampling time section terminates and analyze.Therefore, at the end of each new sampling time section, " Moving Window " is by a sampling time section conversion, such model manager module 914 can consider the data of the nearest acquisition of present sample time period, ignore the data (that is, the oldest data no longer considered) no longer occurred in the time window upgraded simultaneously.Historical time section 1122 can be defined by LTH and LPH, for the present embodiment, after historical time section 1122 follows LTH closely (as shown in Figure 54).LPH also can be called as in this article " nearest historical time section " or " model prediction phase ", and reason becomes obvious in the following description.LTH also can be called as in this article " historical time section remote " or " model training phase ", and reason becomes obvious in the following description.On this point, LTH (historical time section remote) correspond to from train sampling time section 1124 to the time period (comprising endpoint value) terminating training sampling time section 1126, and LPH (nearest historical time section) correspond to from prediction samples time period 1128 to time period (comprising endpoint value) of nearest sampling time section 1120.Therefore, by definition, the present sample time period, (that is, nearest sampling time section 1120) was within the scope of LPH.For this specific embodiment, start the prediction samples time period 1128 corresponding to terminating training sampling time section 1126.Alternatively, start the prediction samples time period 1128 can follow closely terminate training sampling time section 1126 after.

Figure 55 is the process flow diagram of the exemplary embodiment illustrating sensor model manager processes 1150, and process 1150 can be performed by model manager module 914.For ease of understanding, process 1150 shows in the mode of the simplification being conceived to function and describes.The particular formulation of some aspect reference model manager module 914 of process 1150 is in hereafter more detailed description.

Process 1150 represents the circulation performed at current sampling point or time, and it corresponds to nearest sampling time section.This example hypothesis insulin infusion devices has run (task 1150) with to user's body insulin delivery under closed loop mode, and process 1150 receives related data according to predetermined scheme (such as, the sampling time section of five minutes).Therefore, process 1150 receives, and obtains or access the multiple input (task 1154) affecting model manager module 914 and run.Such as, process 1150 can accept at least following data of present sample time period: the current insulin-delivering data representing the amount of the insulin sent by insulin infusion devices in nearest sampling time section process; Represent the current sensor data of the current sensor dextrose equivalent of user, it corresponds to nearest sampling time section; And may need for compensating recently based on the current sensor correction factor of the correction of metering.Any amount of historical data (if necessary) also can be received in task 1154 process.Therefore, the redundancy (it passes for leakage, and the solution of packet loss etc. may be desirable) of some quantity can be built in system.Sensing data can be received in any suitable form and process.Such as, continuous glucose sensor can produce Isig (electric current) value, and it can reflect sensor dextrose equivalent.Model manager module 914 can be appropriately configured into directly process Isig value or model manager module and original I sig value can be changed or be mapped as the expression-form of any expectation.

Process 1150 also can access or retrieve the historical data (task 1156) that sampling time section in the past receives.Task 1156 can represent start-up routine, and it inserts grid, matrix or other forms of database structure as required, thus for hereafter more describe in detail various different calculating, analysis and function preparation model manager module 914.Should be understood that, the following cycle (it performs in the mode of carrying out continuously in closed loop mode) of process 1150 does not need the startup of repetition historical data.And task 1156 can regulate historical data to reflect the new data received simply.For embodiment described herein, following historical data can be processed by model manager module 914: the historical insulin-delivering data of user; And the historical sensor dextrose equivalent of user; But be not limited thereto.Historical insulin-delivering data may correspond to the amount in the insulin sent by insulin infusion devices in each target histories sampling time section process, and historical sensor dextrose equivalent may correspond to the representational sensor glucose measurements in obtaining in each target histories sampling time section process.In some embodiments, each historical sensor dextrose equivalent can be relevant with the sensor calibration factor to history Isig value or obtain from history Isig value and the sensor calibration factor are derivative.

Process 1150 itself is capable of circulation, and the data (see Figure 54) relevant to the historical time section of definition are considered in each circulation.Therefore, process 1150 definable is used for model training phase and the model prediction phase (task 1158) of historical time section.On this point, task 1158 identifiable design or specify which data sample to fall into the model training phase (LTH of Figure 54) and/or which data sample falls into the model prediction phase (LPH of Figure 54).Task 1158 also can be used for identifying or specify " out-of-date " data sample not needing when carrying out forward to consider.In practical operation, if the data of the oldest sampling time section are missed because of some reasons, so process 1150 can make suitable adjustment (such as, search for nearest data available sample, wait for next sampling time section, etc.).

Next, process 1150 processes at least some in historical data to determine the optimum matching scheme (task 1160) of sensor glucose predictions model.Task 1160 can be considered to training program, and it finds best-fit sensor glucose predictions function, this function and then can be used for the integrality and the quality that check (prediction) glucose sensor.In some embodiments, the model tormulation of sensor glucose predictions is quadravalence ordinary differential formula, and this formula is when given starting condition solves, and supply a model the sensor dextrose equivalent predicted.It should be noted that task 1160 be used in the model training phase obtain real sensor dextrose equivalent (and be not used in the model prediction phase obtain any real sensor dextrose equivalent) determine which kind of optional program can be selected as optimum matching scheme.Conceptually, task 1160 produces multiple curve (or can be used for for illustrative purposes making the discrete value of curve visibility) and is compared by the real sensor dextrose equivalent of a part for the curve in the model training phase and the interim acquisition of model training.Have perfect matching ideally, one in the curve of generation can real sensor dextrose equivalent accurately in trace model training period.But in practical operation, the curve produced will depart from real sensor dextrose equivalent.Therefore, task 1160 identifies the curve of the calculating of optimum matching real sensor value.Should be understood that, this optimum matching curve also comprises the sensor dextrose equivalent of model prediction, and it extends beyond the model training phase and enters the model prediction phase.

Process 1150 compares with the corresponding prediction sensor dextrose equivalent of at least one optimum matching scheme by least one the historical sensor dextrose equivalent obtained the model prediction phase and proceeds (task 1162).In some embodiments, task 1162 only checks a real sensor dextrose equivalent: the current sensor dextrose equivalent of sampling time section acquisition recently.In other embodiments, the model prediction phase obtain any sensor dextrose equivalent or all the sensors dextrose equivalent can be analyzed in task 1162 process.Simple embodiment described herein only considers current sensor dextrose equivalent, relatively simple in such task 1162.On this point, (namely task 1162 can calculate current sensor dextrose equivalent, nearest history value) and the current glucose values of the recently prediction of sampling time section between difference (difference can be expressed as absolute value), and task 1150 is by comparing the difference of calculating and threshold error value and proceed (query task 1164).In other embodiments, the comparison performed in task 1162 can comprise more advanced method, such as, considers the curve of the more than one sampled point that model prediction is interim, statistical analysis, etc.Such as, be different from the error of calculation based on point-to-point, process 1150 can use the historical sensor dextrose equivalent in any suitable method Confirming model time span of forecast whether to depart from the model predication value of the correspondence in (at least threshold quantity) optimum matching scheme.

If the error calculated between the actual history sensor dextrose equivalent of the dextrose equivalent of model prediction and correspondence is less than or equal to error threshold, or meet the preassigned that model manager module 914 is monitored, so follow the "No" branch of query task 1164 and process 1150 continues next sampling time section (task 1166).At this moment, process 1150 gets back to task 1152, and the core of such process 1150 can be repeated to consider the data that next sampling time section receives.Therefore, the most legacy data that the previous loops of process 1150 is considered is left in the basket, and the new data received are designated as " recently " data and the historical time section of the previous cycle of process 1150 or " analysis window " produce the displacement (see Figure 54) of a sampling time section.

If the error calculated exceedes threshold error amount (the "Yes" branch of query task 1164), so process 1150 can give the alarm, warn and/or information (task 1168).In practical operation, alarm, warning or information start by model manager module 914, thus carry out translating, circulate a notice of, send, playback, etc.Such as, alarm can be present in insulin infusion devices, remote sensing station, hand held controller equipment, etc.In some embodiments, when exceeding threshold error amount (task 1170), process 1150 is converted to open loop mode (or some types be converted to security operating mode that the insulin with reduction sends) from closed loop mode.

An importance of process 1150 relates to the mode (see task 1160) selecting optimum matching sensor glucose predictions model.On this point, Figure 56 is the process flow diagram of the exemplary embodiment illustrating sensor model training process 1180, and process 1180 can combine execution with the sensor model management process 1150 described in Figure 55.Process 1180 is with understandable simplified way display and describe.The particular implementation of some aspect reference model manager module 914 of process 1180 describes hereinafter in further detail.

As previously mentioned, exemplary sensor glucose predictions model used herein is expressed as quadravalence ordinary differential formula.According to Conventional mathematical method, the real time sensor dextrose equivalent (G) of model prediction is calculated as the function of two model prediction starting condition G0 and dG0.At this, G0 is the sensor dextrose equivalent of the estimation of initial training sampling period 1124 (beginning of the LTH in Figure 54), and dG0 is the derivative of G0.Therefore, different initial conditions values produces the scheme of different sensor glucose predictions models, and each different setting of initial conditions corresponds to different forecast models.For the object for the treatment of effeciency, restriction and boundary are given initial conditions value with the manageable quantity of the stand-by scheme of computation and analysis by model manager module 914.On this point, sensor model training process 1180 starts by the scope or boundary (task 1182) calculating each bounded initial conditions.

For exemplary embodiment provided herein, initial conditions dG ₀to demarcate based on the simple mode of preset parameter (it is adjustable): dG ₀=± grad_bound.On the contrary, initial conditions G ₀border based on the baseline historical sensor dextrose equivalent obtained in the model training phase (or be subject to obtain in the model training phase the impact of baseline historical sensor dextrose equivalent), such as, at the sensor dextrose equivalent that initial training sampling period 1124 obtains.Therefore, process 1180 identifies by historical sensor dextrose equivalent and calculates initial conditions G ₀border object use baseline sensor dextrose equivalent: G ₀=SG _k-LTH± 0.14SG _k-LTHwherein, SG _k-LTHit is the baseline sensor dextrose equivalent (see Figure 54) obtained in the sampling time section the earliest under analysis condition in historical time section.It should be noted that G ₀border be the function of baseline sensor dextrose equivalent, it can change in the mode of carrying out continuously in system operation, and it can change with change of each circulation of process 1180.In practical operation, if sensor glucose data is ignored in initial training sampling period 1124, so process 1180 can be sampled suitable measurement, such as, searches for immediate available sensors glucose data point, waits for next sampling time section, etc.

Process 1180 proceeds (task 1184) by determining, calculating or obtain next group initial conditions subsequently.The mode that process 1180 is undertaken selecting and developing by different initial conditions is unimportant in this article.When last group of initial conditions are used for the stand-by scheme (task 1186) of calculating sensor glucose predictions model.As mentioned above, each stand-by scheme is calculated as the function of two bounded initial conditions.And each stand-by scheme is calculated as the function of user's plasma insulin of estimation, this function and then be calculated as the function of amount of the insulin being delivered to user.Therefore, task 1186 can based on current insulin delivering data (nearest sampling time section obtains); The plasma insulin of the basal insulin velocity estimation user of historical insulin delivering data and user.In practical operation, task 1186 considers total insulin (basis, injects the insulin sent with any other) of whole sampling time section.This allows process 1180 at least partly based on the plasma insulin of estimation and the stand-by scheme obtaining sensor glucose predictions model under analysis condition at the baseline sensor dextrose equivalent of sampling time section acquisition the earliest.

Process 1180 by producing training error value, the function of quantity or stand-by scheme and proceeding (task 1188).Training error calculates from the sensor dextrose equivalent of the prediction of stand-by scheme and the historical sensor dextrose equivalent of correspondence by comparing, thus obtains the tolerance representing predicted value and actual value tight fit.In some embodiments, training error is only based on predicted value and the actual value (LTH in Figure 54) of model training phase, and therefore, task 1188 does not consider any predicted value or the actual value (LPH in Figure 54) of model prediction phase.

If process 1180 has considered the combination (the "Yes" branch of query task 1190) of all initial conditions, so process 1180 can carry out task 1192.If still there are more groups of initial conditions (the "No" branch of query task 1190), so process 1180 can return task 1184, again obtains next group initial conditions and proceeds as mentioned above.Task 1192, after multiple different stand-by scheme is by calculating, uses different group initial conditions to perform.Task 1192 can perform as follows: from the scheme of multiple calculating, select the stand-by scheme of optimum matching.For this particular implementation, select the training error produced in task based access control 118 process.Such as, the stand-by scheme with minimum training error can be selected as optimum matching scheme.

Should be understood that, process 1180 does not need to perform with illustrational order, and some tasks can executed in parallel.Such as, after all stand-by schemes have obtained and preserved, the calculating (task 1188) performing training error can be replaced.And process 1180 can be designed to eliminate immediately the stand-by scheme (after task 1188 completes) that training error exceedes predictive error threshold value.Alternately, if relevant training error meets a certain standard, process 1180 can be designed to specify the stand-by scheme as optimum matching scheme immediately.

Above-mentioned concept and methodology can perform in the actual embodiment of model manager module 914.Following description relates to two kinds of possible embodiments, and it implements generic concept provided above.Should be understood that, particular implementation is hereinafter described not exhaustive, is not intended to the description of embodiment the scope and the application that limit or limit theme as herein described.

Model manager module: first represents embodiment

Model manager module 914 is designed appropriately and is configured to detect incipient fault measurement value sensor.Model manager module 914 can use the mathematical model of off-line training.Such as, can be included, but are not limited to by the parameter that off-line is estimated: K ₁(insulin gain, mg/dL/U/h); τ ₁(the first insulin time constant, minute); τ ₂(the second insulin time constant, minute); Ibasal (basal insulin, U/h); And the SGbase (blood sugar (BG) under fasted conditions, mg/dL) when sending Ibasal insulin.

Model manager module 914 training pattern prediction initial conditions, the G in each sampling time ₀and dG ₀.G ₀and dG ₀represent the BG (mg/dL) under k-LTH condition and BG derivative (mg/dL/min) estimated value (see Figure 54), wherein, LTH is the length in training data (sampling time), and k equals the current sampling time and deducts LPH.In this case, LPH is the length of estimation range in the sampling time.G ₀and dG ₀estimation with demarcating like that of being set forth by the formula of hereafter formula 77 Unified Expression.It should be noted that the task 1182 of these initial conditions and border thereof also reference sensor model training process 1180 as above describes.

G ₀=CGM _k-LTH± 0.14CGM _k-LTH(formula 77)

dG ₀＝±grad_bound

For formula 77, CGM _k-LTHbe the CGM measured value under sampling time k-LTH, and grad_bound is the real-time absolute maximum BG derivative (mg/dL/min) defined in advance.

Model manager module 914 uses from the plasma insulin I under the insulin historical record estimation k-LTH of k-LTH-insulin history and k-LTH (see Figure 54) according to formula 81 _p.From present-LTH-LPH until present, (as described in the task 1186 of sensor model training process 1180 above) generation has the I of estimation _p, G ₀and dG ₀model prediction.Model prediction can calculate two value: Terr and Perr.Terr be defined as model prediction and from k-LTH and k CGM record between mean square of error and (formula 78).Perr be defined as model prediction and from k and present CGM record between absolute average error (formula 79).Should be noted that, Terr is the training error of a type, it is that the task 1188 of process 1180 above describes, and Perr is the predicated error of a type, described by the task 1162 of sensor model management process 1150 above and query task 1164.At Perr < err1 and Terr > err2 time, failure definition (formula 80).

$\mathrm{Terr}=\sqrt{\frac{{\mathrm{\&Sigma;}}_{i=k-\mathrm{LTH}}^k{({\mathrm{Model}}_i-{\mathrm{CGM}}_i)}^2}{\mathrm{LTH}}}$ (formula 78)

$\mathrm{Perr}=\mathrm{abs}\left(\frac{{\mathrm{\&Sigma;}}_{i=k}^{k+\mathrm{LPH}}{\mathrm{Model}}_i-{\mathrm{CGM}}_i}{\mathrm{LPH}}\right)$ (formula 79)

$\mathrm{Fault}=\left\{\begin{array}{cc}1,& \mathrm{if\; Terr}<\mathrm{err}1\mathrm{and\; Perr}>\mathrm{err}2\\ 0,& \mathrm{else\; if\; Terr}\&le;\mathrm{err}2\mathrm{or\; Perr}\&GreaterEqual;\mathrm{err}1\\ -1,& \mathrm{if\; not\; enough\; data\; records\; available}\end{array}\right.$ (formula 80)

In formula 80, Fault 1 represents fault sensor, and Fault 0 represents non-faulting sensor, and Fault-1 represents do not have enough information to determine.The "Yes" branch of query task 1164 is corresponded to reference to Figure 55, Fault 1.

In some embodiments, some in the parameter used by model manager module 914 can be adjustable.Table 5 identifies some exemplary values of adjustable parameter and these parameters.

The adjustable parameter of table 5. model manager module

Following formula describes the mathematical model formula in Laplace transform form:

${\hat{I}}_p\left(s\right)=\frac{1}{(50s+1)(70s+1)}({\hat{I}}_D+{\hat{I}}_{P0}\mathrm{s\&alpha;\&alpha;}+{\mathrm{dI}}_{P0}\mathrm{\&beta;})$ (formula 81)

In this formula, α=3500, β=120, i _pderivative form. and dI _p0be respectively with derivative starting condition.

All insulin situations are expressed in the derivative form from given insulin levels Ibasal, represented by following formula 82:

${\hat{I}}_x=I_x-\mathrm{Ibasal}$ (formula 82)

In formula 82, x represents D, in or P.

Following formula 83 is expressed from the BG in the derivative form of SGbase.It should be noted that a kind of suitable expression formula of formula 83 representative sensor glucose predictions model, it is quadravalence ordinary differential formula.

$\begin{array}{c}\hat{G}\left(s\right)=\frac{1}{({\mathrm{\&tau;}}_{1}s+1)\&CenterDot;({\mathrm{\&tau;}}_{2}s+1)}(K_I\&CenterDot;{\hat{I}}_P-\frac{1}{(50s+1)(70s+1)}({\hat{G}}_0\&CenterDot;\mathrm{\&alpha;}+({\hat{G}}_0\&CenterDot;s+d{\hat{G}}_0)\&CenterDot;\mathrm{\&beta;}+\\ ({\hat{G}}_0\&CenterDot;s^2+d{\hat{G}}_0\&CenterDot;s)\&CenterDot;\mathrm{\&chi;}+({\hat{G}}_0\&CenterDot;s^3+d{\hat{G}}_0\&CenterDot;s^2)\&CenterDot;\mathrm{\&delta;}))\end{array}$

(formula 83)

In formula 83, there is following relationship:

α＝120+τ ₁+τ ₂

β＝3500+120τ ₁+120τ ₂+τ ₁τ ₂

χ＝3500τ ₁+3500τ ₂+120τ ₁τ ₂

δ＝3500τ ₁τ ₂

And, in formula 78, k _i, dG ₀, τ ₁and τ ₂from the derivative starting condition of the BG starting condition in the BG in the derivative form of SGbase, insulin gain, derivative form, BG and two time constants respectively.

Model manager module: second represents embodiment

According to some embodiments, the function of model manager module 914 can be expressed as follows.As mentioned above, model manager module 914 is based on the concentration of glucose of the insulin sent, sensor Isig value and sensor calibration factor real-time estimation user.If the sensor dextrose equivalent (SG) of model prediction and actual SG value are significantly different, system can trigger and represent that the packet collected is containing the failure safe alarm of unaccounted behavior, this failure safe alarm and then to send with fault sensor and/or insulin or the meal ingestion of failing to give notice is associated.

As shown in Figure 54 such of the time range of model manager module 914 and reference time section defines.The method performed by model manager module 914 uses the packet of time in the past range reception to estimate the glucose of plasma insulin and model prediction, thus estimation fault condition.Sampling time is the time interval between two continuous data bags, and for this specific embodiment, the described time interval is five minutes.Insulin history in Figure 54, corresponding to the time in the past scope defined, needs to use this time in the past scope to estimate plasma insulin (for this example, insulin history corresponds to four hours or 48 sampling time sections).The length (LTH) of the training area for this example comprises 24 packets, and it corresponds to the time in the past scope of 120 minutes.The length (LPH) of the estimation range for this example comprises 24 packets, and it corresponds to the time in the past scope of 120 minutes.In Figure 54, the current quantity that k equals packet deducts LPH, and " present " represents the nearest sampling time.

Following formula describes the mathematical model in Laplace transform form.Formula 84 provides the estimation of plasma insulin and formula 85 provides the SG value of model prediction.Therefore, model manager module 914 estimates plasma insulin as follows according to this particular implementation:

${\hat{I}}_p\left(s\right)=\frac{1}{(50s+1)(70s+1)}({\hat{I}}_D+{\hat{I}}_{P0}\mathrm{s\&epsiv;}+{\mathrm{dI}}_{P0}\mathrm{\&gamma;})$ (formula 84)

For the present embodiment, ε=3500, γ=120, be the plasma insulin of the estimation of derivative form, (s) refers to Laplace transform form, and the insulin sent by system of derivative form.And, the plasma insulin (see Figure 54) of the estimation of derivative form in the sampling time being expressed as k-LTH, dI _p0be the derivative of the plasma insulin of estimation, and α and β is constant.

Above-described insulin state is expressed with derivative form, expressed by following formula by given insulin levels Ibasal:

${\hat{I}}_x=I_x-\mathrm{Ibasal}$ (formula 85)

In formula 85, x represents D or P (wherein, D refers to the insulin sent, and P refers to plasma insulin), and I _{basal, 0}the basal rate to the estimation that each user defines, with to patient's imparting value FBG ₀(mg/dL) fasting blood-glucose (FBG).

For this second embodiment, the real time sensor dextrose equivalent of model prediction calculate according to formula 83 and correlationship formula, described by the first embodiment of model manager module 914.On this point, from FBG ₀derivative form model prediction SG value (at the end of night use metering glucose readings estimation blood sugar), (s) refers to Laplace transform form, τ ₁with two insulin time constants that τ is each patient identification, it is relevant with insulin rapid-action degree to patient, K _iinsulin increment, it is the plasma insulin of the estimation of derivative form.And, calculate according to following formula 86, the SG value (mg/dL) of the estimation of derivative form in the sampling time (see Figure 54) of k-LTH, and dG ₀the derivative (mg/dL/min) of the SG value estimated in the sampling time that (being calculated by following formula 87) is k-LTH.Constant α, β, χ, and δ calculates as listed in formula 83 above.

The blood glucose value of estimation is calculated as model prediction initial conditions G ₀and dG ₀function.For this particular implementation, G ₀and dG ₀defining like that of being represented by following formula of estimation.It should be noted that also the task 1182 of reference sensor model training process 1180 is described above for these initial conditions and their border.

G ₀=SG _k-LTH± 0.14SG _k-LTH(formula 86)

DG ₀=± grad_bound (formula 87)

At this, G ₀the SG value (mg/dL) estimated in the sampling time being k-LTH, SG _k-LTHthe SG measured value in the sampling time of k-LTH, dG ₀the derivative of the SG value estimated in the sampling time being k-LTH, grad_bound is predefined absolute maximum SG time-derivative (mg/dL/min).For some embodiments, gran_bound is preset parameter.For embodiment provided herein, the value of grad_bound is 5mg/dL/min.

Model prediction is conducive to the calculating of two value Terr and Perr.Terr is defined as the mean absolute error between the SG value of model prediction and the actual SG record being identified as the sampling time of k-LTH and k of following formula 88 calculating.Perr is defined as the mean absolute error between the SG value of model prediction and the actual SG record being identified as the sampling time of k value present (see Figure 54) of following formula 89 calculating.

$\mathrm{Terr}=\frac{{\mathrm{\&Sigma;}}_{i=k-\mathrm{LTH}}^k\mathrm{abs}({\mathrm{Model}}_i-{\mathrm{SG}}_i)}{\mathrm{LTH}}$ (formula 88)

At this, Terr is defined as the SG value (Model of model prediction _i) and be identified as k-LTH and k sampling time SG record (SG _i) between mean absolute error.

$\mathrm{Perr}=\frac{\mathrm{abs}({\mathrm{Model}}_{\mathrm{present}}-{\mathrm{SG}}_{\mathrm{present}})}{{\mathrm{SG}}_{\mathrm{present}}}\&CenterDot;100\%$ (formula 89)

At this, Perr is defined as the number percent of the error between the SG measured value in model prediction and current (recently) sampling time.

According to this specific embodiment, model manager module 914 estimates failure condition based on formula 90, wherein, Fault 1 represents fault sensor, Fault 0 represents non-faulting sensor, and Fault3 represents training error, and Fault-1 represents does not have enough data to be used for making decision.

$\mathrm{Fault}=\left\{\begin{array}{cc}1,& \mathrm{if\; Terr}<\mathrm{err}1\mathrm{and\; Perr}>\mathrm{err}2\\ 0,& \mathrm{if\; Perr}\&le;\mathrm{err}2\mathrm{or\; Terr}\&GreaterEqual;\mathrm{err}1\\ 3,& \mathrm{if\; Terr}>\mathrm{err}3\\ -1,& \mathrm{if\; not\; enough\; data\; available}\end{array}\right.$ (formula 90)

In formula 90, err1 is the upper limit threshold of mean absolute error (see formula 88).Therefore, if training error is higher than this threshold value, because suspect the reliability of training and not trigger fault.Err2 is the lower threshold of formula 89.If the predicted value of model and current measurement value sensor are higher than this threshold value and training error is less than err1, so can trigger fault.Err3 defines the lower threshold of training time section.If formula 88 represents the value higher than this threshold value, so the alarm relevant with bad training can be triggered.

Figure 57 is the figure of the exemplary sensor condition illustrated corresponding to non-faulting sensor (Fault 0) and fault sensor (Fault 1).Mean level axle represents the current sampling time at rightmost, and represents the time period by LPH and LTH.The most legacy data that sampling time 1202 corresponds to that current time drag manager module 914 considers.Therefore, the historical data 1204 in the sampling time occurred before the sampling time 1202 is ignored.

Top curve 1206 in Figure 57 represents non-faulting sensor (Fault 0), intermediate curve 1208 represents fault sensor (Fault 1), bottom curve 1210 describes the insulin used, and needs to use bottom curve to estimate plasma insulin and the SG value of production model prediction.In curve 1206,1208, solid line 1212 represents the SG value of model prediction, and round dot represents actual SG measured value.Vertical dotted line 1214 represents the boundary between LTH time range and LPH time range.Line between solid line 1212 and round dot represents the difference (error) between the SG value of model prediction and actual SG measured value.Dotted line is used for LPH time range, and it corresponds to 15 minutes or three sampling time sections with regard to this embodiment.

See top curve 1206, between the SG value (being represented by solid line 1212) of model prediction and actual SG measured value (being represented by round dot), there is good consistance.In other words, actual measured value significantly can not depart from predicted value.In some embodiments, model manager module 914 only compares the actual measured value in LPH time range.According to a kind of exemplary embodiment, model manager module 914 only determines malfunction based on the data obtained recently (that is, the information of last sampling time reception).For this example that Figure 55 describes, Perr is less than or equal to err2.Therefore, according to formula 90, model manager module 914 returns Fault 1 and system is instructed to maintain closed loop mode.

See the intermediate curve 1208 of Figure 57, between the actual SG measured value of the SG value of model prediction and LTH time period interior (within this time period, Terr is less than the err1 in formula 90), there is good consistance.However, it is noted that, be finally worth at the SG of model prediction and observe significant difference between the last SG measured value 1218 (Perr is greater than the err2 in formula 90) of reality.Therefore, in this case, model manager module 914 can send failure safe alarm and/or carry out other suitable measurements.

In some embodiments, module management module 914 use parameter in some can be adjustable.Table 6 illustrates some adjustable parameters of this embodiment, and some exemplary values of these parameters.

Parameter	Default value	Lower limit	The upper limit
				K I(mg/dL/U/H)	-100	-360	-49
FBG 0(mg/dL)	120	50	300
				Ibasal(U/H)	1	0.1	3
err1(mg/dL)	5	1	30
				err2(％)	50	20	100
err3(mg/dL)	10	1	30
				LTH (sampling time)	24	4	48
LPH (sampling time)	24	1	48

The adjustable parameter of table 6. model manager module

Leak transmission module

Leak transmission module 916 continuous review controller whether receive packet (comprising SG value) for the treatment of.When not receiving the packet (such as, a row is less than four packets, represents that time span is less than the packet sum of 15 minutes, etc.) being less than defined amount, leaking transmission module 916 and maintaining system and run under closed loop mode.In this time course, system uses closed loop control algorithm to continue to calculate insulin dose based on last effective sensor dextrose equivalent or sensor Isig value.For expression higher than lower limit time threshold and higher than upper limit time threshold value (such as, 15 minutes to 60 minutes) the packet do not received, leak transmission module 916 system is converted to the foundation for security speed of programming in advance, its can be defined as patient night basal rate half.If start to receive packet in foundation for security Velocity Time scope internal controller, so system converts back to closed loop mode.But for the packet do not received represented higher than the time of upper limit time threshold value, leak transmission module 916 and system is converted to open loop mode to send the basal rate at night of programming in advance, it can be set by health care providers or paramedic.

Leak transmission module 916 and check following different situations: which kind of type the packet lost in transmitting procedure belongs to and when lost data packets in transmitting procedure.Different steps based on lose transmission type and perform.The details of four kinds of different situations describes hereinafter.

Case 1

If sensor Isig value and SG value are all received by controller, so:

A sensor Isig preserved by () controller;

B SG value preserved by () controller;

C zero-order holder (ZOH) counting is set to 0 by (); And

D () as previously mentioned, system remains on closed loop mode.

Case 2

If sensor Isig value is not received, but controller have received SG value, so:

A () ZOH counting is set to 0;

B () uses SG value and the sensor calibration factor to calculate Isig by formula 91 (vide infra); And

C () system remains on closed loop mode.

Isig _calc=(SG/CF ')+2 (formula 91)

Case 3

If have received sensor Isig value but controller does not receive SG value, so:

A ZOH counting is set to 0 by ();

B () uses Isig value and the sensor calibration factor to calculate SG by formula 92 (vide infra); And

C () system held is at closed loop mode.

SG _calc=(Isig-2) × CF ' (formula 92)

Case 4a

If controller had not both received sensor Isig value and had not received SG value (that is, two values all do not receive) yet, and if:

ZOH Count≤ZOH Count Max

So:

A the ZOH counting of () sensor Isig and SG calculates based on preceding value;

(b)ZOH Count＝ZOH Count+1；

(c) TimeoutCount=0; And

D () system held is at closed loop mode.

Case 4b

If controller had not both received sensor Isig value and had not received SG value (that is, two values all do not receive) yet, and if:

ZOH Count＞ZOH Count Max

So:

A () preserves the engineering noise placeholder of sensor Isig value and SG value;

B () system held at closed loop mode, but is converted to interim foundation for security speed, this speed is the half of patient at night basal rate when open loop mode;

If c () system acceptance sends foundation for security speed to packet simultaneously, so system can convert back to closed loop mode;

(d) for system send foundation for security speed per minute for, TimeoutCount progressively increases: TimeoutCount=TimeoutCount+1;

If e () TimeoutCount>TimeoutCount is Max, so system is converted to open loop mode.

According to some embodiments, although different values can suitably for particular implementation, ZOH Count Max has fixed value 2, and Timeout Count Max has fixed value 45.And the foundation for security speed of leaking transmission module 916 use can be adjustable.On this point, foundation for security speed can about 0 to 5 unit/hour scope in regulate.

Although provide the embodiment that at least one is exemplary in the detailed description above, should be understood that, but also there is a large amount of change.Should be understood that, exemplary embodiment described herein or numerous embodiments limit the scope of claimed theme, application or structure unintentionally by any way.And detailed description above can be the convenient road map figure that those skilled in the art provide embodiment described by enforcement or numerous embodiments.Should be understood that, under the condition not deviating from claim limited range, various different change can be made in the function of element and arrangement, these change be included in submit to present patent application time known equivalent and forseeable equivalent.

Claims (33)

1. control the method performed by processor of the insulin infusion devices of user, described method comprises:

Run the processor structure that comprises at least one processor device with what obtains biologically active insulin estimated value in representative of consumer body and work as insulin (IOB) value on header board;

IOB speed is calculated by described processor structure at least partly based on obtained current I OB value;

Determined the infusion of insulin speed regulated at least partly by described processor structure based on the IOB speed calculated and uncompensated infusion of insulin speed; And

The final infusion of insulin speed of described insulin infusion devices is selected by described processor structure, wherein, the described infusion of insulin speed selecting determined adjustment, uncompensated infusion of insulin speed or current basal speed are as final infusion of insulin speed.

2. the method for claim 1, described method also comprises:

The insulin of described insulin infusion devices is regulated to send according to selected final infusion of insulin speed.

3. method as claimed in claim 2, described method also comprises:

Selected final infusion of insulin speed is transferred to described insulin infusion devices.

4. the method for claim 1, described method also comprises:

Rerun according to predetermined scheme, calculate, determine and select, thus regulate final infusion of insulin speed in the mode of carrying out continuously.

5. the method for claim 1, wherein described in move to small part and inject delivering data based on the history of user and obtain current I OB value.

6. method as claimed in claim 5, wherein, described operation obtains current I OB value according to three Room Insulin Pharmacokinetics models.

7. the method for claim 1, wherein calculate IOB speed to comprise:

When the IOB value obtained is greater than minimum IOB value, the IOB value that calculating IOB speed equals to obtain is multiplied by the IOB rate of decay; And

When the IOB value obtained is less than or equal to minimum IOB value, calculates IOB speed and equal 0.

8. the method for claim 1, wherein determine that the infusion of insulin speed regulated comprises:

According to expression formula AdjustedRate (n)=max (0; PIDRate (n) – IOBRate (n)) select regulate infusion of insulin speed, wherein:

AdjustedRate (n) is the infusion of insulin speed of selected adjustment;

PIDRate (n) is uncompensated infusion of insulin speed; And

IOBRate (n) is the IOB speed calculated.

9. the method for claim 1, wherein select described final infusion of insulin speed to comprise:

When PIDRate>Basal according to expression formula FinalRate (n)=max (Basal; AdjustedRate (n)) select described final infusion of insulin speed, wherein:

FinalRate (n) is selected final infusion of insulin speed;

Basal is current basal speed;

AdjustedRate (n) is the infusion of insulin speed of the adjustment determined; And

PIDRate (n) is uncompensated infusion of insulin speed.

10. the method for claim 1, wherein select final infusion of insulin speed to comprise:

When described current basal speed is more than or equal to uncompensated infusion of insulin speed, final infusion of insulin speed is selected to equal uncompensated infusion of insulin speed.

11. the method for claim 1, described method also comprises:

Uncompensated infusion of insulin speed is calculated by described processor structure according to proportional, integral-derivative insulin feedback (PID-IFB) control algolithm.

12. 1 kinds for controlling the method performed by processor of the insulin infusion devices of user, described method comprises:

By the processor structure comprising at least one processor device produce the estimated value of the biologically active insulin in representative of consumer body when insulin (IOB) value on header board;

IOB speed is settled accounts by described processor structure at least partly based on produced current I OB value;

Obtain uncompensated infusion of insulin speed;

According to expression formula AdjustedRate (n)=max (0; PIDRate (n) – IOBRate (n)) by described processor structure determine regulate infusion of insulin speed; And

According to expression formula $\mathrm{FinalRate}\left(n\right)=\left\{\begin{array}{cc}\max (\mathrm{Basal};\mathrm{AdjustedRate}\left(n\right)),& \mathrm{PIDRate}>\mathrm{Basal}\\ \mathrm{PIDRate}\left(n\right),& \mathrm{PIDRate}\&le;\mathrm{Basal}\end{array}\right.$ Select final infusion of insulin speed;

Wherein:

AdjustedRate (n) is the infusion of insulin speed of the adjustment determined;

PIDRate (n) is the uncompensated infusion of insulin speed obtained;

IOBRate (n) is the IOB speed calculated;

FinalRate (n) is the final infusion of insulin speed selected; And

Basal is the current basal speed maintained by the insulin infusion devices of user.

13. methods as claimed in claim 12, described method also comprises:

Selected final infusion of insulin speed is transferred to described insulin infusion devices to be sent by the insulin of described insulin infusion devices to promote to regulate.

14. methods as claimed in claim 12, described method also comprises:

Repeat to produce, calculate, obtain, determine and select according to predetermined scheme, thus regulate final infusion of insulin speed in the mode of carrying out continuously.

15. methods as claimed in claim 12, wherein, described generation injects delivering data calculating current I OB value based on the history of user at least partly.

16. methods as claimed in claim 12, wherein, described generation calculates current I OB value according to three Room Insulin Pharmacokinetics models.

17. methods as claimed in claim 12, wherein, described calculating calculates IOB speed according to following expression formula:

$\mathrm{IOBRate}\left(n\right)=\left\{\begin{array}{cc}\mathrm{GainIOB}\&times;\mathrm{IOB}\left(n\right),& \mathrm{IOB}\left(n\right)>\mathrm{MinIOB}\\ 0,& \mathrm{IOB}\left(n\right)\&le;\mathrm{MinIOB}\end{array}\right.;$

Wherein,

GainIOB is the IOB rate of decay; And

MinIOB is minimum IOB value.

18. methods as claimed in claim 12, wherein, described acquisition comprises:

PIDRate (n) is calculated by described processor structure according to proportional, integral-derivative insulin feedback (PID-IFB) control algolithm.

19. 1 kinds for controlling the method performed by processor of the insulin infusion devices of user, described method comprises:

Described insulin infusion devices is run with insulin delivery to user's body in closed loop mode;

Obtain the insulin data of the current delivery of the amount representing the insulin sent by described insulin infusion devices in nearest sampling time section;

Obtain the current sensor data of the current sensor dextrose equivalent representing the user corresponding to nearest sampling time section;

Process insulin data and historical sensor data that the history in the multiple history samples time periods before nearest sampling time section sends, thus obtain the sensor dextrose equivalent of the prediction of historical time section;

Calculate the difference between current sensor dextrose equivalent and the current sensor dextrose equivalent of prediction in nearest sampling time section, wherein, the sensor dextrose equivalent of the prediction in historical time section comprises the current sensor dextrose equivalent of prediction; And

Give the alarm when described difference exceedes threshold error value.

20. methods as claimed in claim 19, described method also comprises:

Response gives the alarm, and closed loop mode is converted to open loop mode.

21. methods as claimed in claim 19, wherein:

Nearest historical time section correspond to from the prediction samples time period to the time period of nearest sampling time section, comprise endpoint value;

Farthest historical time section correspond to from train sampling time section to terminate training sampling time section time period, comprise endpoint value;

The described section of historical time farthest occurred before nearest historical time section; And

Described process obtains as the sensor dextrose equivalent of the model prediction of the function of bounded starting condition, the impact of the baseline sensor dextrose equivalent obtained in historical time section farthest described in described bounded starting condition is subject to.

22. methods as claimed in claim 21, wherein, the described beginning prediction samples time period corresponds to described end training sampling time section.

23. methods as claimed in claim 21, wherein, described baseline sensor dextrose equivalent is starting to train sampling time section to obtain.

24. methods as claimed in claim 19, wherein, described process comprises:

Based on the insulin data of described current delivery, the plasma insulin of the basal insulin velocity estimation user of the insulin data that described history is sent and user; And

At least partly based on the sensor dextrose equivalent of the plasma insulin acquisition model prediction of estimation.

25. methods as claimed in claim 19, wherein, described process comprises:

Calculate the stand-by scheme of multiple sensor glucose predictions model,

Wherein, each in described multiple stand-by scheme is calculated as the function of bounded starting condition, and wherein, described bounded starting condition is subject to the impact of baseline sensor dextrose equivalent, and described historical sensor data comprises described baseline sensor dextrose equivalent; And

From the calculated stand-by scheme of multiple users, select optimum matching scheme as the sensor dextrose equivalent of model prediction.

26. 1 kinds of users control the method performed by processor of the insulin infusion devices of user, and described method comprises:

Described insulin infusion devices is run with insulin delivery to user's body in closed loop mode;

The baseline historical sensor dextrose equivalent of training sampling time section to obtain being identified in from the historical sensor dextrose equivalent of user;

Calculate the stand-by scheme of multiple sensor glucose predictions model, wherein, the function of the insulin data that each history being calculated as bounded starting condition and user in described multiple stand-by scheme is sent, and wherein, described bounded starting condition is subject to the impact of described baseline sensor dextrose equivalent;

From calculated multiple stand-by scheme, optimum matching scheme is selected with comparing of the Part I of described historical sensor dextrose equivalent based on from the sensor dextrose equivalent of the prediction in calculated multiple stand-by scheme;

Relatively from the sensor dextrose equivalent of at least one prediction and the Part II of described historical sensor dextrose equivalent of described optimum matching scheme, wherein, the Part I of described historical sensor dextrose equivalent corresponds to historical time section farthest, the Part II of described historical sensor dextrose equivalent corresponds to nearest historical time section, and before the described section of historical time farthest occurs in the nearest historical time section of carrying out data sampling; And

When the Part II of described historical sensor dextrose equivalent depart from described optimum matching scheme at least threshold error amount time, response ratio comparatively gives the alarm.

27. methods as claimed in claim 26, described method also comprises:

Response gives the alarm, and is converted to open loop mode from closed loop mode.

28. methods as claimed in claim 26, wherein, described calculating comprises:

The plasma insulin of the basal insulin velocity estimation user based at least some in the insulin data that the history of user is sent and based on user, wherein, each in described multiple stand-by scheme calculates based on the plasma insulin of estimation at least partly.

29. 1 kinds of users control the method performed by processor of the insulin infusion devices of user, and described method comprises:

Described insulin equipment is run with insulin delivery to user's body in closed loop mode;

Find the optimum matching scheme of the sensor glucose predictions model of the historical sensor dextrose equivalent obtained relative to the model training phase, wherein, described optimum matching scheme is the function of the baseline sensor dextrose equivalent that the model training phase obtains and is the function of the insulin data that the history of user that historical time section obtains is sent;

Relatively from sensor dextrose equivalent and at least one the historical sensor dextrose equivalent only corresponding to the model prediction phase of at least one prediction of described optimum matching scheme; And

Sensor dextrose equivalent that at least one historical sensor dextrose equivalent described departs from least one prediction described at least threshold error amount time, response ratio comparatively gives the alarm.

30. methods as claimed in claim 29, described method also comprises:

Response gives the alarm, and is converted to open loop mode from closed loop mode.

31. methods as claimed in claim 29, wherein, nearest historical sensor dextrose equivalent and the nearest prediction sensor dextrose equivalent from optimum matching scheme compare by described comparison.

32. methods as claimed in claim 29, wherein, after the described model prediction phase follows the model training phase closely.

33. methods as claimed in claim 29, wherein, described in find optimum matching scheme to comprise:

Calculate the multiple stand-by scheme of described sensor glucose model;

Relatively from the prediction of the multiple stand-by scheme calculated sensor dextrose equivalent with only at the corresponding historical sensor dextrose equivalent that the model training phase obtains, thus each respective training error of the multiple stand-by scheme of acquisition calculating; And

From the multiple stand-by scheme calculated, optimum matching scheme is selected based on described training error.